The

change
issue
Volume 124 Jul/Aug USD $9.99
Number 4 2021 CAD $10.99
Technology, trends, At EmTech MIT, immerse yourself in bold
thinking and visionary innovation on the most
and the new rules significant technologies and global trends that
of business will propel business and shape our world.

JOIN US LIVE ONLINE

September 28-30, 2021

 MEET
S T
U

PL

H
E
MIT Technology Review’s flagship event Innovators
on emerging technology and trends Under 35
OF 1
202

AT EMTECH MIT, WE’LL EXPLORE:

The widespread effect The global impacts The new rules of

of intersecting of technology business for a world that
technologies such as consolidation, mandates corporate
AI, IoT, biotech, and misinformation, energy, social responsibility
cybersecurity and climate

SUBSCRIBERS SAVE 10% WITH CODE PRINTJA21 AT

EmTechMIT.com/RegisterNow
 02 From the editor

The power of us

T
hough we’ve called it the “Change” issue, this edi-
tion of the magazine is really about two things:
reflection and empowerment.
For far too many of us, the pandemic has been
a study in feeling powerless, and we’ve had little
time to reflect as we focus on keeping ourselves
and our loved ones safe, employed, and as mentally sound as
possible. We’ve been forced to cope almost
constantly with the twisting, morphing
uncertainties that life has thrown at us.
And yet in this unprecedented envi-
ronment incredible stories of hope and
empowerment have emerged. We see peo-
ple finding ways to respond to suffering
and injustice with positive change. Take
the stories Abby Ohlheiser has collected
(page 9), including those of Carlisa Johnson,
who turned a Google Doc into a nexus of
power for the Black Lives Matter move-
ment, and Fiona Lowenstein, who nur-
tured an online community of thousands
into a place where those suffering from
covid-19 can get vital information. Sarah
Jaffe writes (page 74) that a failed vote to
unionize Amazon workers at a facility in
Alabama may be discouraging, but around
the US, workers in the increasingly expan-
sive tech sector are waking up to their
power to organize, and to demand dignity.
In an essay on the arc of progress (page Michael Reilly dangerously misleading, if for no other
is executive
16), Sheila Jasanoff harks back to West reason than that they can be interpreted
editor of
Bengal in India, where she was born, and MIT Technology to justify individualism at all costs. In the
tells how under British rule the region’s Review US, this attitude has been corrosive to sup-
thriving industry of woven textiles was port for government funding of important
crushed by the Industrial Revolution. The high-tech industries like chip fabrication,
lesson isn’t that technological advancement is bad—it’s that we which, as Jeremy Hsu writes, is one reason America is racing to
must take care not to assume that all such change is for the best, catch up to manufacturers overseas (page 54). We have similar
or that it comes without costs. work to do in the rapidly evolving field of clean energy, where—
As Jasanoff writes, the good news is that we are not bystanders as Gernot Wagner writes—the price of solar panels has tumbled
in the process. We are the ones who create technology, after all; over the last few decades (page 82). With a bit of a boost from
we have the power to choose what gets built and how it is used. further R&D funding and favorable policies, solar stands a real
Nowhere is this agency on fuller display than in this year’s list chance of helping decarbonize the planet.
of 35 Innovators Under 35 (page 23). I hope you’ll take time to Reflection can lead to positive change, but it needs to come
sit with this list. I find it impossible not to come away inspired with empowerment. As Karen Hao notes in her feature (page
by their accomplishments—from swarms of French-toast size 48), a group of marginalized AI workers suffered indignities
satellites to new research into fusion power to a pair of budding for years in the white-male-dominated field before finding one
companies racing to bring optical computing to market. These other. With that sense of community, they realized they had the
innovators are literally creating the future before our eyes. power to challenge the biggest companies in the world to be
As we know, each of them stands on the achievements of better and more inclusive.
SIMON SIMARD

those who have come before. And yet the tech world is replete If nothing else, I hope when you’re through with this issue,
with narratives about single-minded mavericks bucking ortho- you’ll have given yourself time to reflect, feel your own power,
doxy to realize their vision of the future. Those stories can be and just maybe find that you’re a little bit changed.
“First Republic took the time to get to know me —
that real human connection is everything.”
BLAIR HOLBROOK
Operations Strategy

(855) 886-4824 | Ŕrstrepublic.com | New York Stock Exchange symbol: FRC

MEMBER FDIC AND EQUAL HOUSING LENDER
04
Report
9 23
Activists’ online toolbox
These organizers used apps
and social media to energize
others during the pandemic.
By Abby Ohlheiser
14
The era of autonomy
Zoox CEO Aicha Evans talks
about leading the self-driving-
car company and the power
of opportunity.
By Anthony Green
16
Four arguments
Sheila Jasanoff discusses the
dangerous appeal of tech.
Joe Garcia asks how
prisoners can reenter a world
transformed by digital tech.
Jennifer Neda John reflects
on the ways online influencers
can mislead.
Kiara Royer ponders whether
online activism can really lead
to change.
Contents
THE CHANGE ISSUE
The list Features Review
48 74
35 Innovators under 35 The fight to reclaim AI Books, culture, and
Our annual list of 35 rising How a small group of policy in perspective
young innovators is an oppor- researchers at the margins Sarah Jaffe on how to have
tunity to take a look not just at built a movement to wrest a labor movement in the age
where technology is now but the world’s most powerful of Big Tech.
where it’s going and who’s tak- emerging technology from
James Surowiecki on the
ing it there. Big Tech’s control.
task facing modern-day
By Karen Hao
trustbusters.
+ Inventors . . . . . . . . . . . . . . . . . . .24
54 Gernot Wagner on what cheap
+ Humanitarians . . . . . . . . . . . 30
The great chip divide solar power means for climate
+ Visionaries . . . . . . . . . . . . . . . . .34
A shortage of microchips is change.
+ Pioneers . . . . . . . . . . . . . . . . . . . .38
threatening to slow digital
+ Entrepreneurs . . . . . . . . . . . . .44 Wudan Yan on the toll
innovation inspired by
of online harassment.
the promise of ever faster,
cheaper computing power.
By Jeremy Hsu The back page
60
88
Those who fall behind The change chronicles
get beaten up
Can science transcend
nationalism?
By Yangyang Cheng
66
Fast time
A series of photographs taken
around the world show how
climate change is warping
geological time.
Photos by Ian van Coller
Cover illustration
by Sophy Hollington
But wait,
there’s
more.
Lots more.
You’re already a subscriber. • Unlimited web access
Activate your account and • Exclusive digital stories
start enjoying: • The Algorithm newsletter
• Access to 120+ years of
publication archives

technologyreview.com/subonly
06 Masthead

Editorial Corporate Consumer marketing MIT Technology Review

Executive editor Chief executive officer and publisher Senior vice president, Insights and international
Michael Reilly Elizabeth Bramson-Boudreau marketing and consumer revenue Vice president, Insights and international
Doreen Adger Nicola Crepaldi
Editor at large
David Rotman Assistant to the CEO Director of digital marketing Director of custom content, US
Katie McLean Emily Baillieu
News editor Laurel Ruma
Niall Firth Human resources manager Director of event marketing Senior project manager
James Wall Brenda Noiseux
Managing editor Martha Leibs
Timothy Maher Manager of information technology Growth marketing manager Content manager
Colby Wheeler Em Okrepkie
Commissioning editors Jason Sparapani
Bobbie Johnson Office manager Assistant consumer marketing manager Senior manager of licensing
Konstantin Kakaes Linda Cardinal Caroline da Cunha Ted Hu
Amy Nordrum
Circulation and print production manager Director of custom content, international
Senior editor, MIT News Product development Tim Borton Francesca Fanshawe
Alice Dragoon Chief technology officer Email marketing manager Director of business development, Asia
Senior editor, biomedicine Drake Martinet Tuong-Chau Cai Marcus Ulvne
Antonio Regalado Director of software engineering
Senior editor, climate and energy Molly Frey Advertising sales Board of directors
James Temple Head of product Vice president, sales and Martin A. Schmidt, Chair
Senior editor, digital culture Mariya Sitnova brand partnerships Peter J. Caruso II, Esq.
Abby Ohlheiser Andrew Hendler
Senior project manager Whitney Espich
Senior editor, cybersecurity Allison Chase andrew.hendler@technologyreview.com Jerome I. Friedman
Patrick Howell O’Neill 646-520-6981 David Schmittlein
Software engineer
Executive director, integrated marketing Glen Shor
Senior editors, AI Jack Burns
Caitlin Bergmann Alan Spoon
Karen Hao
Will Douglas Heaven caitlin.bergmann@technologyreview.com
Events Customer service and
Senior editor, computing Executive director, brand partnerships
Siobhan Roberts
Senior vice president, Marii Sebahar subscription inquiries
events and strategic partnerships marii@technologyreview.com
Senior editor, podcasts and Amy Lammers 415-416-9140 National
live journalism 877-479-6505
Director of event content Executive director, brand partnerships
Jennifer Strong International
and experiences Kristin Ingram
Podcast producer Brian Bryson kristin.ingram@technologyreview.com 847-559-7313
Anthony Green 415-509-1910 Email
Head of international and custom events
Editor, Pandemic Technology Project Marcy Rizzo Senior director, brand partnerships customer-service@technologyreview.com
Lindsay Muscato Debbie Hanley Web
Senior event content producer
Senior reporters Kristen Kavanaugh debbie.hanley@technologyreview.com www.technologyreview.com/
Tanya Basu (humans and technology) 214-282-2727 customerservice
Event content producer
Eileen Guo (technology policy Senior director, brand partnerships MIT Records (alums only)
Erin Underwood
and ethics) Ian Keller 617-253-8270
Associate director of events ian.keller@technologyreview.com
Reporters Reprints
Nicole Silva 203-858-3396
Charlotte Jee (news) techreview@wrightsmedia.com
Neel Patel (space) Event operations manager Senior director, brand partnerships 877-652-5295
Tate Ryan-Mosley (data and audio) Elana Wilner Miles Weiner
miles.weiner@technologyreview.com Licensing and permissions
Copy chief Event partnership coordinator licensing@technologyreview.com
Linda Lowenthal Madeleine Frasca 917-275-2048

Social media editor Digital sales strategy manager

Events associate
Benji Rosen Casey Sullivan
Bo Richardson MIT Technology Review
casey.sullivan@technologyreview.com
Engagement editor 617-475-8066 One Main Street
Abby Ivory-Ganja Finance Media kit 13th Floor
Audience engagement associate Finance director www.technologyreview.com/media Cambridge, MA 02142
Madison Umina Enejda Xheblati 617-475-8000
Proofreader General ledger manager The mission of MIT Technology Review
Barbara Wallraff Olivia Male is to make technology a greater force for
good by bringing about better-informed,
Accountant more conscious technology decisions
Design Anduela Tabaku through authoritative, influential, and
Chief creative officer trustworthy journalism.
Eric Mongeon Technology Review, Inc., is an
Art director independent nonprofit 501(c)(3)
Emily Luong corporation wholly owned by MIT;
the views expressed in our publications
Marketing and events designer and at our events are not always shared
Kyle Thomas Hemingway by the Institute.
Photo editor
Stephanie Arnett
CANCER DOESN’T STOP.
NEITHER DO WE.
FOR ONE NIGHT. WE STAND TOGETHER.

SATURDAY 8/21
8 ET/7 CENTRAL

American Lung Association’s LUNG FORCE, Amgen, Cless Family Foundation, Fanconi Anemia Research Fund, Farrah Fawcett Foundation,
Jazz Pharmaceuticals, Laura Ziskin Family Trust, Legacy Circle, LUNGevity Foundation, Mirati Therapeutics, Pancreatic Cancer Canada,
Sara Schottenstein Foundation, Society for Immunotherapy of Cancer, Lew, Jean, and Kari Wolff
Stand Up To Cancer is a division of the Entertainment Industry Foundation (EIF), a 501(c)(3) charitable organization.
 A DV E R T I S E M E N T

The Green Future Index 2021

ENERGY TR ANSITION

KEY

O Green leaders
O Greening middle
O Climate laggards
O Climate abstainers

Top 5 Bottom 5
Rank Country Score Rank Country Score
Q The energy transition pillar measures the
1. . . . . . . ........... Ethiopia ....... . . . . . . . . . . . . . . . . . . 8.1 76 . . . . . . . . . . . . . . . Qatar .............................. 1.8 degree to which each country or territory
is promoting renewable energy.
2 . . . . . . ............ Angola ......... . . . . . . . . . . . . . . . . . . 7.4 75 . . . . . . . . . . . . . . . Ukraine .......................... 1.9 Q Seven of the top 10 countries are in
Africa, leading the way for their recent
3 . . . . . . ............ Uganda ........ . . . . . . . . . . . . . . . . . . 7.4 74. . . . . . . . . . . . . . . . Russia ............................ 2 rate of change.

Q The political will to implement an effective

4 . . . . . . . ........... Cameroon . . . . . . . . . . . . . . . . . . . . 7.4 73 . . . . . . . . . . . . . . . Iran .................................. 2
energy transition has been absent in Hong
Kong, and the territory is highly dependent
5 . . . . . . ........... Nigeria ......... . . . . . . . . . . . . . . . . . . 7.3 72 . . . . . . . . . . . . . . . Hong Kong.................... 2.1 on coal-fired power.

The Green Future Index is a ranking of 76 countries The index ranks the green performance of countries
and territories building a low-carbon future. It mea- and territories across five pillars:
sures how economies are pivoting toward clean energy, • Carbon emissions • Clean innovation
industry, agriculture, and society through investment in • Energy transition • Climate policy
renewables, innovation, and green finance. • Green society

Experience the interactive index, view

the data, and download the full report at: technologyreview.com/gfi
The Green Future Index was produced in association with Salesforce, Citrix, and Morgan Stanley.
 09

RE
We wanted to think about the root cause
of this, not as singular incidents ... but as the
product of structural, systemic racism.
—PAGE 12

VO
P I ERT
W
LearnBooks,
how people
are using
arts, andtechnology
culture
to bring about social and
in perspective
political change.

CARLISA JOHNSON, 29
GEORGIA

Johnson created a widely shared Google

Doc with links to educational material,
contact details for public officials, ways to
take action, and information on the Black
Lives Matter movement in the days after
George Floyd was murdered by a Minne-
apolis police officer in May 2020. The doc-
ument became a resource for activists,
particularly those new to the movement.

uring last summer’s itera-

ACTIVISTS’ D tion of Black Lives

Matter, a lot of people
were entering activism for the
first time. On the personal side,

ONLINE TOOLBOX we were in the midst of a pan-

demic and I live with someone
who is very immune compro-
mised, so I couldn’t go to protests,
which is what I usually do. And I
How four organizers used apps and social media to energize others
felt very helpless. So I needed to
during the pandemic. By Abby Ohlheiser
figure out a way that I could feel
For the past year, people in the United States were supposed to live at a like I was actually having some
distance and keep to themselves to prevent the spread of a deadly virus. At sort of impact.
the same time, there was a nationwide reckoning on systemic racism and The people who are directly
inequality, a contentious election, and a rise in racist violence. The activ- around me, those who I inter-
GRACE HEEJUNG KIM; COURTESY PHOTO

ism that met those crises was often organized wholly or partially online, act with most on social media—
sometimes by young people experimenting with creative new ways to build they are ingrained in the world of
social movements. We spoke to four people in their 20s who, through the activism in the same way that I
internet, became viral voices, key organizers, and vital resources over the am. There weren’t necessarily pri-
past year. Their stories are shared here in their own words and have been vate conversations where people
edited and condensed for clarity. were having these eye-opening
moments or reckonings, where
they were like, “Oh my gosh,
10
something needs to be done. I
don’t know what to do.” The peo-
ple I’m in community with—they
know what to do.
This document was something,
I felt, for my friends to share out
with their family members and
their friends. That specifically
speaks to me as a Black person.
Most Black people didn’t need to
have these conversations because
they already know this. And a lot
of my community is Black people.
I have a lot of academics as
friends. I created my post, and
my call to actions, on my personal
Facebook and made it public, and
then my friends shared it. That’s
how it snowballed. It went from
campus to campus. I have no clue
how this happened, but it started
to go to celebrities as well. So the
cast of [the TV series] Riverdale
started sharing it. I noticed that
there were a lot of teenagers ask-
ing if it was okay to share it, which
is a demographic that I have no
access to.
It’s kind of cliché now, but this
activism, working toward correct-
ing inequities—you have to oper-
ate as a chorus. When one voice
goes out, there are others who
are still sustaining. And so even
though my document was live
during that specific time period
when so much was happening, it
was happening so fast, and I just
couldn’t sustain it toward the end.
But there are so many other docu-
ments that exist that have created
this network that is still flourish-
ing today. Q
I needed to figure
out a way that I could
feel like I was actually
having some sort of
impact.
Change
FIONA LOWENSTEIN, 27
Fiona
Lowenstein’s NEW YORK
Slack group
for people with
covid-19 now has Lowenstein is the founder and editor in
chief of Body Politic, a media organiza-
10,000 tion and wellness collective based in New
MEMBERS York that hosts a Slack support group for
people with covid-19, including those with
and gets about
long-term symptoms. It now has more than
50 to 100 new
members each 10,000 members and gets about 50 to 100
week. new members each week.
The group was the meeting place for
the Patient-Led Research Collaborative, a
global collective of covid-19 patients who
are recording and sharing data about their
own symptoms, and which has begun to
publish research on long-term covid.
got sick very early on
I when the pandemic hit
the United States. My
first symptoms were on March
13, 2020. I was sick before there
was a comprehensive symptom
Julie 8:09 PM
Today I walked 30 minutes without getting tired, hugged my
partner for the first time since mid-March, frosted the birthday
cake he made me, and just found myself in the kitchen ... of all
things ... whistling! Day 53
list from the CDC, before there
was any information about long-
term recoveries or young people
getting severely sick. Those first
couple of weeks that I was sick
and in the hospital, I just lacked a
lot of information.
The support group started
as an emotional support group.
People with covid and people
with long covid really needed
a place to talk to each other.
But then it quickly became an
info-sharing group, because we
were lacking information from
our doctors and from health
agencies. We were just sort of
talking to each other and trying
to figure it out ourselves. The
group was on Instagram actu-
ally at the time, as a DM chat. It
had maybe, like, 25 to 30 peo-
ple in it.
A lot of people with myalgic
encephalomyelitis, or chronic
fatigue syndrome, reached out
to us kind of early on in the
pandemic. They’re, for the
most part, people who have
long-term symptoms following
GRACE HEEJUNG KIM; COURTESY PHOTOS

viral infections. They reached
out to provide guidance on how
to navigate an illness like this
but also on health, advocacy,
and how to interact with health
agencies.
I think that really informed
the way that we moved forward,
and made sure that we were
always contextualizing long
covid within these broader long-
term illnesses and chronic ill-
nesses. This is not entirely new.
There’s actually a whole history
of this stuff.
Our support group is really
just for covid patients, but we do
have a special advocacy chan-
nel where we have some users
from other chronic illness and
disability and health justice orga-
nizations that talk about some of
these more intersectional issues.
For those experiencing or
recovering from covid-19, we
have channels for almost every
system of the body: reproduc-
tive, neurological, muscular, cir-
culatory, gastrointestinal. That’s
where we’ll discuss very specific
symptoms. We also have chan-
nels for specific communities.
We have three private chan-
nels that you join by request:
a BIPOC channel, an LGBTQ
channel, and a channel for medi-
cal professionals.
And then there are chan-
nels that are a little bit more
geared toward people’s specific
mental-health needs. We have a
Victories channel. That’s where
you post everything from “I took
a shower for the first time in a
week” to “I reunited with my
family after six months.” We also
have a channel called Need to
Vent, which is kind of the oppo-
site—it’s where you go when you
COURTESY IMAGE
just need to really spill your guts
on how you’re feeling and how
things are going. Q
Report 11
ERYNN CHAMBERS, 28
NORTH CAROLINA
Chambers saw some bad statistics about
crime in Black neighborhoods being used
to support and spread racist narratives on
TikTok—and, having started building a fol-
lowing on the app, decided to debunk them
in song:
Black neighborhoods are
overpoliced, so of course they
have higher rates of crime.
And white perpetrators are
undercharged, so of course
they have lower rates of crime.
Erynn
And all those stupid stats you
Chambers I use TikTok knowing that it’s
wrote a song keep using are operating off a
not the ideal platform. It’s where
to debunk small sample size.
statistics about
I have the most followers. But
crime in Black So shut up! Shut up, shut up, it’s been pretty good. I’ve met
neighborhoods shut up, shut up, shut up, shut up! a lot of people that are friends
that eventually now through TikTok. I made a
had The song rocketed around the Inter-
lot of connections, which has
net and was shared widely beyond TikTok
TWO last summer, eventually earning more
been really good, really encour-
aging. Honestly, that’s probably
MILLION than 2 million views. Chambers—
@Rynnstar on TikTok—has continued
the biggest reason why I still do
VIEWS to be active and popular on the app, with
720,000 followers.
it, because a lot of times it’s emo-
tionally exhausting.
on TikTok.
TikTok kind of pigeonholes
a lot of Black creators into “You
was no great inspiration can only talk about this one
I T or anything like that. I thing.” It’s frustrating how TikTok
was just standing on my seems to pick and choose what
porch one day last summer and I they let be seen on the platform.
sang a random tune that popped I really try to promote and push
into my head and hit Post. The when I do other things, because
next day, it had really taken off, I want people to know that I’m
and I was stunned. And then more than just political videos
another creator, Alex Engelberg, they share with their aunts. I’ve
made a remix, a choral remix of got a lot of interests. I want to
it—basically a barbershop quar- be able to talk about my random
tet. That really made it take off. hobbies as much as any white
Maybe it was just the right creator. I also enjoy talking about
place, right time, you know? People linguistics. I like talking about
found it snappy, especially with the musical theater. I like talking
barbershop remix. Very few people about history. And I don’t want
plan to go viral, but everything that people to think that if they follow
I’ve done that has gone viral was me for one thing, they’re never
completely unexpected. going to see anything else. Q
12 Change

immigrants. And so they’re largely

SUNNIE LIU, 22
TEXAS
In the summer of 2020, Sunnie Liu, an
undergraduate student at Yale, was
part of a small group of young Chinese-
Americans who wanted to find ways
to address anti-Blackness within their
own community. So they co-founded the
WeChat Project.
WeChat, a Chinese app that is kind of
like a social network, messaging service,
and sharing app in one, is extremely pop-
ular with the Chinese diaspora in the US.
The project creates content that seeks to
counteract what is often overwhelmingly
right-wing discourse, news, and misinfor-
mation shared there. So far, the project has
published more than 25 bilingual articles
that have reached hundreds of thousands
of readers across social media.
here’s not that many
T young people of the
Chinese diaspora on
WeChat. And there’s very few
progressive voices on there. So
even though conservatives are a
minority among Chinese-
Americans and Asian-Americans
in general, right-wing discourse
and political information really
dominate the WeChat platform.
The way that WeChat works is
that political news usually spreads
from media accounts into group
chats. These group chats often
older than us. They’re probably
part of our parents’ or grandpar-
ents’ generations. For the Chinese
diaspora, there’s not only this gen-
erational divide; there’s often also
a language and a cultural divide.
At the dinner table, you have par-
ents who are most comfortable
speaking Mandarin, and the kids
are most comfortable speaking
English. And so the kid could be
trying to express some compli-
cated political idea in English,
but then the parents have no idea
what’s going on, and vice versa
when the parents are speaking
Mandarin.
They have very different trau-
mas that they’ve experienced in
either generation, too, with the kids
growing up and experiencing rac-
ism, perhaps since they were five
and entered into American schools, The WeChat parents, and the grandparents and
versus the parents growing up in Project is the grandchildren. They’re curious
trying to
China, where there’s not really what their children are thinking.
the same racism. But they might
have lived through the Cultural
BRIDGE In response to anti-Asian
violence, for us, especially on
Revolution. We’re trying to bridge THE GAP WeChat but also in general—hon-
that gap by starting these conver- by starting estly, throughout the news media,
sations between the generations conversations we felt like most coverage of the
between the
online in a language that meets the topic was missing class analysis
generations
parents and grandparents of the online. and gender analysis. Especially
Chinese diaspora where they are. after Atlanta, this is something
The main source of news that all Asian-Americans need to
on WeChat comes from media worry about. This is an attack on
accounts. WeChat calls them “offi- our lives, on our safety.
cial accounts.” They’re essentially But if you look at who was
microblogs that publish articles actually attacked, the vast major-
people can read and share and ity of these people were the most
discuss, and like and comment on. vulnerable in our communities.
We’re out there publishing with They are low income. They are
some of the rare progressive media elderly. Often, they’re immigrants
accounts on WeChat. who do not speak English. And
GRACE HEEJUNG KIM; COURTESY PHOTO

just end up being echo chambers
for spreading sensationalism,
conspiracy theories, and unfortu-
nately lots of right-wing rhetoric.
The vast majority of the active
WeChat users in the States are
members of the Chinese dias-
pora who are first-generation
Because we’re a rare voice
of the children’s and grandchil-
dren’s generation, a lot of people
in the older generations are curi-
ous because of these cultural and
language barriers. Unfortunately,
there’s a lot of miscommunica-
tion between the kids and their
these are very different from the
people who are most visible and
vocal in response. And so we
wanted to also think about the
root cause of this, not as singular
incidents by racist people, but as
the product of structural, systemic
racism and classism. Q
Active Molecule Unlocked
As we age, our ability to generate vital cellular energy declines. This can lead to
a decline in muscle function, metabolism, energy levels and resiliency. Aging is an
inevitable part of life, but scientists have discovered a molecule that rejuvenates the
power sources inside our cells – our mitochondria.

Clinical studies show that the natural molecule, Urolithin A, improves mitochondrial
function and boosts leg muscle strength*. But, it’s not as simple as eating more
pomegranates. Most people’s microbiomes don’t produce enough Urolithin A by
eating the precursors found in pomegranates. Our scientists in Switzerland have
*See disclaimers on website.

spent more than a decade researching and developing the first clinically tested, highly
pure form of Urolithin A, unlocking its benefits for everyone. We called it .
A precise dose of 500mg Mitopure delivers six times more Urolithin A than a
glass of pomegranate juice.*

timelinenutrition.com/sa
14 Change

T R :

Q + A

The past year has been a wild

ride for Aicha Evans. Zoox,
the autonomous-vehicle com-
pany Evans leads as CEO,
was acquired by Amazon last
summer for a reported $1.2
billion. And when the com-
pany unveiled its vehicle in
December, the car represented
a significant departure from
the automobile as we know it.
Meant to serve as a self-driv-
ing taxi, it looks more like a
high-tech carriage than a car.
Sliding glass doors welcome
passengers from either side,
and each of the vehicle’s outer
corners houses a “sensor pod”
with multiple lidar and radar
modules and cameras to help
it navigate. Beneath the floor-
board is an electric motor that
can whisk passengers to their
destination at up to 75 miles
per hour.
Zoox’s vehicle is one of the
few driverless rigs built from

THE ERA
the ground up, and Evans says
it’s also one you’ll never get to
own. Instead, Zoox plans to
launch an app-based ride-hail-
ing service in cities including

OF AUTONOMY
San Francisco and Las Vegas,
where the vehicle is being
tested.
I spoke with Evans about
what it’s like to try to change
an industry and transform the
way we move about our cities.
Zoox CEO Aicha Evans on
leading the self-driving-car company and the power of opportunity.

By Anthony Green | Portrait by Winni Wintermeyer


Q: How do you define Zoox?
Is it an AI company? A robot-
ics company?
A: It’s a transportation com-
pany that takes advantage of
AI, robotics—all of the new
techniques around electric
vehicles and software—and
blends all of that to basi-
cally open up a new area of
transportation.
A city like San Francisco
that has housing issues and
is worried about business
flight has 30% of its real
estate footprint dedicated to
parking. So if people were
using Zoox to go from point
A to point B, those buildings
could be replaced, reclaimed
for businesses or for housing
and parks.
The other thing that’s
really important from a Zoox
standpoint is taking advan-
tage of sensors and com-
puting to make all of this
happen. One of the questions
we get all the time is: Why
are you building a vehicle?
Well, because the passenger
car of today was architected
and designed for human driv-
ers. Rearchitecting and rede-
signing the vehicle to make
it easiest and safest for AI to
drive is what we’re all about.
Q: How might autonomous
vehicles affect our lives when
we’re not on the road?
A: The world 30, 40 years
from now will look very dif-
ferent. We talk about auton-
omy as really the beginning
of a wave. Sort of like also
what happened with the
internet and then the PC and
then wireless, and then the
smartphone.
I mean, the smartphone is
not that old, right? Sometimes
Report
I’m like, how did we even
operate without these
things? Well, we did. And I
think autonomy will allow
a lot of things like that—
around goods, around ser-
vices. I think there are a lot
of things we physically go
to a place for today that in
the future will come to us
through autonomy.
Q: Many driverless cars are
primarily trained in Western,
urban environments. How well
will these systems work in
other places?
A: Mathematical algorithms
are not biased. However data
can be. Not because the data
is bad but because of where
you collect the data. In terms
of Zoox, what I can tell you
is that we will not go some-
where without training on
local data sets. If you don’t do
that and you make assump-
tions, life could get quite
complicated.
I think that as an indus-
try, we understand the sci-
ence, and it’s important to
understand from an input
standpoint what could be
problematic. You’re also
more likely to do that if there
are people in the room who
don’t all look the same and
think the same.
Q: What’s your approach
to leadership?
A: It’s changed over the years.
You’re first an engineer, and
you get noticed because
you’re one of the best in the
room on the project. And
you try to improve and learn
more and have more impact,
but very quickly you sort of
understand the math around
one versus many. And as a
single person, you can only
do so much.
Leadership is not com-
mand and control. It’s really
about: How do you get peo-
ple together? How do you get
them sold and bought in on
a mission? And then how do
you work together to accom-
plish it?
Q: When it comes to building
teams, what’s your approach
to diversity?
A: We’re building a consumer
product at the end of the day,
and consumers come in all
shapes, all races, all genders,
all … everything. And so it’d
be bad to build a consumer
product without having peo-
ple who look like and think
like the consumers.
Another element of it is
access and equity. Like, for
an African-American woman
with short hair, getting my
hair cut is a big deal. Not a lot
of people can cut my hair—let
me just put it that way. You
start noticing any time you
move to a new city: “Oh, I just
have to find MLK Boulevard.
And on either side of it, all the
barbers will know how to cut
my hair.” And that was true in
DC. That was true in Austin.
That was true in Portland.
And so I look forward to our
vehicle picking up people in
whatever neighborhood they
are from, and basically giving
them the opportunity to be
transported to where the eco-
nomic access is.
Q: You’re a Black woman
in power, in tech. That is not
common. How do you deal
with that?
A: I don’t actually wake up
every day thinking I’m a
15
Black woman with power in
tech. Occasionally when I’m
getting pulled over or when
somebody is doing some-
thing stupid, I am reminded,
but I don’t want to be an
angry person all the time.
I don’t think that’s very
productive.
What I do think about is,
I represent opportunity. I
represent that it is possible.
I’ve spent a lot of time asking
myself: What is embedded
systematically that makes this
rare, and how can we break
that down?
I remember going to a
Lego robotics competition
with my son for the first time
in the Bay Area. And my son
said to me, “Wow, there’s
only one of me here.” I’m like,
“No, no, no, you’re confused.”
And I looked around, and
yes, there was only one of
him and there were very few
girls’ teams. And the lack of
opportunity is already right
in front of you.
And then you bring that to
Zoox and you say, “Hey, crew,
we’re going to sponsor Lego
robotics. Are some of you
interested in being mentors?”
I look at opportunity. I look at
how do you make a meaning-
ful, positive difference?
This is our problem as a
nation. These are problems
that have not been solved for
centuries. Our nation was
built and started a certain
way. So a little bit of honey
and less vinegar might be the
right way to go.
This interview has been con-
densed and edited for clarity.
Anthony Green produces
podcasts at MIT Technology
Review.
16
ARGUMENTS
THE DANGEROUS
APPEAL OF TECH
Technology doesn’t rule us. We direct it, but often by inaction.
By Sheila Jasanoff
ove it or hate it, technology
L
enthralls us with the prom-
ise of change. Sometimes
it’s the presumed benefits
that grab our attention:
curing disease, replacing
fossil fuels, increasing food
supplies, unlocking the
secrets of the deep sea,
colonizing Mars, or end-
ing the ravages of old age.
Other times the risks loom
larger. What if we unleash a killer virus,
set in motion a nuclear doomsday, block
out harmful solar radiation with chemicals
that prove toxic, or build computers that
decide humans are dispensable?
The battle between light and dark in the
way we imagine technological change is
ancient. In Greek mythology, Prometheus
suffered agonies for bringing fire to Earth,
and Daedalus lost his son to the urge to fly
to freedom. But the most optimistic and
Change
most pessimistic views of technology both
rely on a common misconception: that a
technological pathway, once embarked
upon, leads to inevitable social conse-
quences, whether utopian or dystopian.
This view, known as technological
determinism, is historically flawed, polit-
ically dangerous, and ethically question-
able. To achieve progress, societies like
ours need a more dynamic understand-
ing of why technology changes, how we
change with it, and how we might govern
our powerful, marvelous machines.
Technology is not an autonomous force
independent of society, nor are the direc-
tions of technological change fixed by
nature. Technology at its most basic is
toolmaking. Insisting that technological
advances are inevitable keeps us from
acknowledging the disparities of wealth and
power that drive innovation for good or ill.
Technology is always a collective ven-
ture. It is what it is because many people
Sheila Jasanoff
is a professor
of science and
technology
studies at the
Harvard Kennedy
School.
imagined it, labored for it, took risks with
it, standardized and regulated it, van-
quished competitors, and made markets
to advance their visions. If we treat tech-
nology as self-directed, we overlook all
these interlocking contributions, and we
risk distributing the rewards of invention
unfairly. Today, an executive officer of a
successful biotech company can sell stock
worth millions of dollars, while those who
clean the lab or volunteer for clinical trials
gain very little. Ignoring the unequal social
arrangements that produced inventions
tends to reproduce those same inequalities
in the distribution of benefits.
Throughout human history, the desire
for economic gain has underwritten the
search for new tools and instruments—in
fields like mining, fishing, agriculture, and
MARTHA STEWART
recently gene prospecting. These tools open
up new markets and new ways to extract
resources, but what the innovator sees as
progress often brings unwanted change

to communities colonized by imported
technologies and their makers’ ambitions.
For example, in West Bengal, where
I was born, weavers lost such skills as
making the intricate narrative motifs of
the Baluchari sari during 200 years of
British rule. Indeed, Britain’s first industrial
revolution, which introduced the power
loom in cities like Lancaster but adopted
punitive tariffs to keep out hand-loomed
cloth from India, was also a story about
dismantling Bengal’s once-flourishing tex-
tile industry. Lost arts had to be regained
after the British left. The cost of a radical
break with a nation’s own economic and
cultural heritage is incalculable.
The desire for military advantage is
another driver of technological change
that can, in some instances, benefit civil
society—but “dual use” technologies often
SELMAN DESIGN
retain ties to forces that prompted their
development. Nuclear energy, a spinoff
from the pursuit of the atomic bomb,
Report
was sold to the world by US President
Dwight Eisenhower as “atoms for peace.”
Yet nuclear power remains closely tied to
the threat of nuclear weapons proliferation.
Similarly, the internet and world wide
web, which revolutionized how much of
the world lives today, owe much to the US
Defense Department’s vision of a network
of computers. First celebrated as a space for
emancipation, the digital world has slowly
revealed its antidemocratic features: con-
stant surveillance, cybersecurity threats, the
lawlessness of the dark web, and the spread
of misinformation. More public awareness
of the internet’s origins might have led to
a more accountable cyberworld than the
one designed by hotshot technologists.
The story of the internet shows that
modern societies are often better at imag-
ining the upsides of technology than its
downsides. But the trajectory of innovation
is also guided by more subtle cultural pref-
erences, often with profound consequences.
17
In US biomedicine, for example, energy,
attention, and money tend to be directed
to high-impact, silver-bullet solutions,
or “moonshots,” rather than to messier
changes in the social infrastructures that
give rise to many health problems.
This inclination is reflected in
Congress’s decision to authorize $10 bil-
lion for Operation Warp Speed to bring
a covid-19 vaccine quickly to market.
Moderna owes much of its success as
a vaccine manufacturer to that massive
public spending, and both Moderna and
Pfizer have benefited hugely from lucrative
supply contracts with the US government.
At the same time, about a third of all
US deaths from the pandemic occurred
in nursing homes, a result of decades of
underinvestment in the unglamorous social
practices of elder care. Collectively, we
chose to ignore the plight of the vulnera-
ble elderly, and spent big on technology
only when everyone was at risk.
Change may not be inevitable, but
economists have a point when they talk
about “path dependency,” or the notion
that once an engine gets going it’s bound
to follow an existing track. Sunk costs—
foundations laid, machinery ordered,
workforces trained—cannot be recov-
ered. It often seems easier to go where
the flows of materials and social practices
have already cut deep channels. It’s not
surprising, then, that defense spending
has proved to be one of the prime moti-
vators of innovation, even though such
investments perpetuate power imbalances
and seldom respect cultural or ethical
sensitivities.
In his famous poem “The Road Not
Taken,” Robert Frost reflects on how the
human mind constructs narratives of inev-
itability. We come to a fork in the road, we
choose a path, and then as memory plays
its tricks we come to see that choice as
shaping all that came after. Faced with
mounting problems of inequality, dimin-
ishing resources, and a looming climate
calamity, we must learn to recognize the
flaws in such linear storytelling, and to
imagine the future along as-yet-untraveled
pathways of change. Q
18
ARGUMENTS
Joe Garcia is a
correspondent
for the Prison
Journalism
Project and a
staff writer for
San Quentin News
at San Quentin
State Prison,
where he is
incarcerated.
LEFT BEHIND
How are prisoners supposed to reenter society when
technology has moved on?
By Joe Garcia
alifornia recently promised to
C
provide free computer tablets to
all state prisoners by the end of
2021, allowing prisoners like me
to email our loved ones through
a highly restricted prison mes-
saging service and download
content like movies and books.
It’s a great first step, but without more
open and frequent internet access, there’s
no way we’ll ever truly keep pace with the
changing world outside our prison walls.
I’ve been locked up since 2003. Back
then Apple had barely launched iTunes,
and I was still in awe of the so-called high-
speed connection I’d paid Time Warner
to install in my apartment. In all the years
since then, I haven’t logged a single second
of internet activity. My frames of reference
for what it means to be online now come
from network television and print media.
When my first parole opportunity
arrives, in 2023, I’ll be 53 years old. As
Change
a convicted murderer, I will need to con-
vince the parole board not only that I have
been rehabilitated, but also that I am able
to be a productive and employable citizen.
It can be hard to keep up with changes in
technology even when you’re experienc-
ing them firsthand. When you’re locked
away, it’s virtually impossible.
I’m hoping to pursue a career path in
journalism when I get out of prison, and I
worry every day about reentering a global
economy with my grossly outdated tech
skill set. I know the job market will expect
internet fluency more than ever in the post-
pandemic world, since much of America’s
media workforce has gone remote.
Roughly 2.3 million people are incarcer-
ated in the US. Even though the internet
is a given in the rest of society, access in
prison is so restricted it’s almost nonexis-
tent. Prisoners are only allowed to use a tiny
number of programs that might offer Zoom
classes with outside teachers, or to browse
an extremely limited set of whitelisted sites
through intranets that are carefully cor-
doned off from the public internet.
At San Quentin State Prison, where
I reside, the computers that prisoners
can access feature preloaded interactive
programs that offer the same basic expe-
rience as reading from a textbook. My
only experience using a search engine
has been through LexisNexis, which the
prison library licenses to allow us to study
case law.
In San Quentin’s vaunted coding pro-
gram, Code 7370, offered through the
prison education organization The Last
Mile, hand-selected prisoners build and
sell actual websites for commercial use.
But even they do not have internet access.
As a working journalist in prison, I
know I have First Amendment rights, but
I’m deprived of a key technology that the
United Nations identified a decade ago
as a means of exercising one’s freedom
of expression.
I understand why government officials,
prison administrators, and the public might
fear that giving convicted felons real-time
internet access would open a Pandora’s
box of suspicious activity. Yet if elemen-
tary school kids can surf the web safely
with parental locks and controls, how
hard would it be for someone to design a
system that provides incarcerated people
with more meaningful access?
In Belgium, for example, an innovative
platform called PrisonCloud has offered lim-
ited and controlled internet access to pris-
oners for years. In Finland and Denmark,
open prisons, which have minimal security
and some of the lowest recidivism rates,
also allow limited internet access.
PHOTO: EDDIE HERENA; ILLUSTRATION: DANIEL ZENDER
Almost all incarcerated people in US
prisons today will be released back into their
communities in the future. That includes
many of my peers who have been incarcer-
ated since before the advent of the internet.
President Joe Biden has said that the
majority of incarcerated Americans deserve
a bona fide second chance at life—but we
need some form of internet access in order
to have a genuine chance of successfully
reentering today’s tech-driven society. Q

The rise of influencers has made
it easier for young people to fall for
misinformation.
By Jennifer Neda John
IN SOCIAL
MEDIA
WE
TRUST
teenage girl peers gravely at
A
the camera, the frame wobbling
as she angles her phone at her
face. A caption superimposed
on her hoodie shares an omi-
nous warning: If Joe Biden is
elected president of the United
States, “trumpies” will commit
mass murder of LGBT individuals and peo-
ple of color. A second caption announces,
“this really is ww3.” That video was posted
to TikTok on November 2, 2020, and liked
more than 20,000 times. Around that
time, dozens of other young people shared
similar warnings across social media, and
their posts drew hundreds of thousands
of views, likes, and comments.
Clearly, the claims were false. Why,
then, did so many members of Generation
COURTESY PHOTO
Z—a label applied to people aged roughly
9 to 24, who are presumably more digitally
savvy than their predecessors—fall for
such flagrant misinformation?
Report 19
I’ve worked as a research assistant at LGBT, and their past posts discussed famil-
the Stanford Internet Observatory since iar topics like family conflict and strug-
last summer, analyzing the spread of gles in math class. This sense of shared
online misinformation. I’ve studied for- experience made them easy to believe,
eign influence campaigns on social media even though they offered no evidence for
and examined how misinformation about their claims.
the 2020 election and covid-19 vaccines Making matters worse was the infor-
went viral. And I’ve found that young mation overload many people experience
people are more likely to believe and pass on social media, which can lead us to trust
on misinformation if they feel a sense of and share lower-quality information. The
common identity with the person who election rumor appeared among dozens
shared it in the first place. of other posts in teenagers’ TikTok feeds,
Offline, when deciding whose claims leaving them with little time to think
should be trusted and whose should be critically about each claim. Any efforts
ignored or doubted, teenagers are likely to challenge the rumor were relegated to
to draw on the context that their commu- the comments.
nities provide. Social connections and As young people participate in more
individual reputations developed through political discussions online, we can expect
years of shared experiences inform which those who have successfully cultivated
family members, friends, and classmates this identity-based credibility to become
teenagers rely on to form their
opinions and receive updates
de facto community leaders,
attracting like-minded people
on events. In this setting, a
community’s collective knowl-
edge about whom to trust on
which topics contributes more
to credibility than the identity
and steering the conversation.
While that has the potential
to empower marginalized
groups, it also exacerbates
the threat of misinformation.
of the person making a claim, Jennifer Neda People united by identity will
John is a soph-
even if that identity is one the omore at Stan- find themselves vulnerable
young person shares. ford University to misleading narratives that
majoring in
Social media, however, human biology.
target precisely what brings
promotes credibility based on She researches them together.
online misin-
identity rather than commu- Who, then, has a role to play
formation at the
nity. And when trust is built Stanford Inter- in promoting accountability?
on identity, authority shifts to net Observatory. Social media platforms can
influencers. Thanks to looking implement recommendation
and sounding like their followers, influ- algorithms that prioritize a diversity of
encers become trusted messengers on voices and value discourse over clickbait.
topics in which they have no expertise. Journalists must acknowledge that many
According to a survey from Common readers get their news from social media
Sense Media, 60% of teenagers who use posts viewed through the lens of iden-
YouTube to follow current events turn to tity—and present information accordingly.
influencers rather than news organiza- Policymakers must regulate social media
tions. Creators who have built credibility platforms and pass laws to address online
see their claims elevated to the status of misinformation. And educators can teach
facts while subject matter experts strug- students to assess the credibility of sources
gle to gain traction. and their claims.
This, in large part, is how the rumor of Shifting the dynamics of online dia-
plans for post-election violence went viral. logue will not be easy, but the dangers
The individuals who shared the warning misinformation can fuel—and the prom-
were deeply relatable to their audience. ise of better conversations—compel us
Many were people of color and openly to try. Q
20 Change

ARGUMENTS

consulted in social or political matters, even

VIRTUAL POLITICS though digital platforms have provided us

with a voice and a way of expressing it ear-
lier in life (an estimated 81% of teenagers 13
to 17 are now active on at least one social
media site). That may stem from a feeling
that our voices don’t matter because we
cannot vote until we turn 18. But most of
us will be able to by the next presidential
election in 2024, if not sooner.
Digital platforms have the potential
to redefine civic engagement and allow
the opinions of both young and old peo-
ple to play a deeper role in policymaking.
As my generation speaks out online, the
lawmakers who are shaping our future will
need to figure out how best to listen to
those of us who will live in it. Otherwise,
young people’s enthusiasm for politics
could dry up. At a time when our trust in
government is nearing historic lows, the
future of political participation is at stake.

Young people are speaking out online. Will it lead to real-world action? Digital democracy
By Kiara Royer The idea that some combination of tech-
nology and a new generation is redefin-
ing politics is not new—the same thing
happened with the radio, and later with
television. But social media, in particular,
ast summer, my friend Jessica Rosberger texted me has brought unique changes. That means

L
with an idea. “I think I may have something,” she began. my generation has a special role to play in
We were about to graduate from high school and had figuring out how these platforms get used.
spent the last three months of senior year taking classes The ways young people use such tools
at home because of the covid-19 pandemic, and lately are already changing the look of political
we’d been following the news of racial justice protests campaigns and grassroots organizing.
around the US in the wake of George Floyd’s murder. Many nonprofits and other groups are now
Kiara Royer
An hour and a half later, we published Jessica’s idea is a freshman recruiting more and more young people to
as an online petition. In it, we argued that former attorney gen- at Williams play larger roles within their organizations.
College majoring
eral William Barr, who graduated from our high school and was The key to making sure young people
in history
given the Distinguished Alumni Award in 2011, had violated the and political stay engaged is including them in more
school’s core values with his involvement in the violent removal science. She is political conversations, says Beth Simone
the news section
of protesters from Lafayette Square in Washington, DC, on June editor for the Noveck, director of New York University’s
1, 2020. We hoped our petition would encourage the school’s Williams Record. Governance Lab and New Jersey’s first
alumni council to rethink Barr’s award. chief innovation officer. Noveck leads a
Jessica and I coordinated over Google Docs, talked with project called CrowdLaw, which stud-
reporters and alumni over Zoom, and shared the petition on ies ways lawmakers can use technology
Instagram, Facebook, and Twitter. By July, it had more than to incorporate the opinions of citizens,
8,700 signatures, was cited in an op-ed in the Washington Post, especially young ones, into the legislative
and propelled us to a virtual meeting with the alumni council. process. She also heads a GovLab pro-
It was my first taste of the power of using the internet and social gram called ReinventED, which centers
YUKO ROYER

media as political tools. Unfortunately, it’s a feeling that’s still too on using technology to engage students,
rare, even for my generation—young citizens’ opinions are rarely educators, and caregivers, especially from

marginalized communities, in efforts to
solve education issues.
Exercises completed by ReinventED
show that students’ priorities even in the
midst of a pandemic lean toward solving
real-world problems and improving nontra-
ditional academic subjects. Policymakers,
on the other hand, are more concerned with
public health and school reopening plans.
“The people who are most expert in
education—mainly students and teachers,
and to a lesser extent the parents of those
students—are rarely, if ever, consulted in
how we design our schools,” Noveck says.
“My hope is that by using tools like this, by
laying bare what people really care about,
that can help to change the direction of
what we’re focusing on.”
Digital platforms, however, may be a
double-edged sword. Participating in online
ALLIE SULLBERG
movements may not translate into offline
engagement—some experts warn it could
have the opposite effect. “On social media,
Report
you can get a burst of interest, sometimes
a burst of activity, because it’s so easy to
feel like you’ve participated just by clicking
a link or retweeting something or using a
hashtag,” says Nicholas Carr, a sociology
professor at Williams College. “What’s
unclear is whether social media will help or
hurt the ability of activists to sustain inter-
ests in a long-term campaign of change.”
Instead, the result may be “slacktiv-
ism,” a term coined during the rise of the
internet for the practice of publicly sup-
porting a cause in ways that take little
effort, often to make yourself look good.
“That can diminish or even demean the
seriousness of political discourse in a way
that can kind of hinder our ability to solve
big problems,” says Carr.
People who engage in this performa-
tive activism are still spreading political
messages, though, says William Golub, a
junior at Stanford University who volun-
teered with the texting team on Joe Biden’s
21
presidential campaign last year. “I think
that there certainly are people who will
just post about something on social media
and that’s the end of the chain, but lots of
those people are people who wouldn’t have
done anything at all,” he says.
Staying engaged
After we met with the alumni council
last July, months passed, and Jessica and
I hadn’t received any updates on Barr’s
award. Frustrated, we published an open
letter to the council on Medium in early
September. The council responded two
days later with a public update stating that
it would share its decision once its written
report was complete.
“Ultimately,” read the report released
two months later, “we would not recom-
mend revoking the Award bestowed on
then-former Attorney General Barr in 2011.”
(Barr held the post from 1991 to 1993, and
again in 2019–2020.) The council said this
decision was based on community feed-
back, the “complex” process of revocation
and precedent, and the lack of “undisputed
information available” regarding Barr’s
involvement at Lafayette Square.
It was devastating. I felt as though the
council, whose youngest member gradu-
ated from high school in 2002, had dis-
missed our efforts.
And yet, I can see now that our work
wasn’t in vain. Our school’s student-run
newspaper published an in-depth analysis
of the report, admonishing the council for
its decision. Jessica and I received emails
from our former teachers, who said our
petition had sparked classroom discussions
about topics ranging from Barr’s actions
to political engagement more broadly.
And the council contacted Jessica and me
directly, thanking us “for taking an active
role in alumni affairs and for your early
dedication, as alumnae, to the legacy of
the school.”
Even though the final decision was not
what I had hoped for, the experience taught
me that my voice is just as important as
the voices of people much older, and that
technology can help make it heard. But
people must be willing to listen. Q
Artificial
intelligence,
demystified
The Algorithm Exclusively for
newsletter MIT Technology Review subscribers.

technologyreview.com/algorithm

I N N O V
A T O R S
U N D E R
The 35 Innovators Under 35 is our yearly opportunity
to take a look at not just where technology is now,
but where it’s going and who’s taking it there. More
than 500 people are nominated every year, and from
this group the editors pick the most promising 100
to move on to the semifinalist round. Their work is
23
then evaluated by our panel of judges (you can see
who they are on page 47), who have expertise in
such areas as artificial intelligence, biotechnology,
software, energy, and materials. With the insight
gained from these rankings, the editors pick the
final list of 35. —The editors
 24
Change

GUTTER CREDIT HERE

 35 Innovators Under 35
Their innovations
point the way
toward light-based
chips, better gene
editing, and skin-
like electronics.
RYAN BABBUSH
AGE AFFILIATION
32 Google
IN N O V A T I O N
Efficient quantum
simulation algorithms
might help find novel,
powerful materials.
Molecules are complicated. Forget
the grade-school picture of electrons
orbiting a nucleus like planets around
the sun. Electrons can be shared
among many atomic nuclei. They
interact with one another in ways
described by the equations of quan-
tum mechanics. It’s these complex
interactions, which grow exponen-
tially with the number of electrons,
that largely govern chemical reactions
and the properties of molecules.
Simulating these electrons with
perfect precision might take a con-
ventional computer millions of years.
But algorithms running on quantum
RYAN YOUNG
computers might be able to perform
precise computations in days or even
hours. This would provide clues on
how to precisely design molecules
with desired properties and tailor
their reactions with amazing control.
Sufficiently precise quantum
simulation might allow chemists to
create new compounds like better
high-temperature superconduct-
ors, catalysts that could take nitro-
gen or carbon dioxide out of the air,
new drugs, more efficient solar cells,
strong lightweight materials for air-
planes, and so forth. It would be a way
to quickly figure out how a new sub-
stance would behave without actually
having to synthesize it. It might her-
ald a new age of materials science.
Between 2014 and 2020, Ryan
Babbush published dozens of
papers—together with collaborators
at Google and elsewhere—that out-
lined dramatically more efficient
quantum simulation algorithms. The
upshot is that some quantum simula-
tion calculations could, in principle,
be done in hours, on a sufficiently
powerful quantum computer.
Take the case of nitrogenase, an
enzyme that some bacteria use to
25
remove nitrogen from the air and
create ammonia, a compound of
nitrogen and hydrogen. This pro-
cess, known as nitrogen fixation,
is essential for agriculture, which is
why nitrogen-based fertilizers are a
linchpin of the world’s food system.
Nitrogenase is a big molecule that
includes a catalytic site known as
FeMoco.
Currently, an energy-intensive
technique known as the Haber-Bosch
process produces most fertilizers,
accounting for about 2% of human-
ity’s total energy usage. “If we could
figure out how that enzyme [nitro-
genase] is doing this, then we might
be able to design an industrially viable
alternative for producing fertilizer,
which could scale and save a huge
amount of energy,” Babbush says.
He and his collaborators have
found a potential way to use a quan-
tum computer to analyze FeMoco
and shed light on the mechanism
by which it first breaks the bonds
between nitrogen atoms that are
bound together in nitrogen gas and
then succeeds in combining the
nitrogen with hydrogen. (Babbush
acknowledges that competing
approaches using clever approxima-
tions to simulate molecules on clas-
sical computers might get there first.)
Another line of research that
Babbush has advanced aims to fig-
ure out how quantum computers can
calculate the behavior of electrons in
metals and crystals. Potential appli-
cations could include finding better
superconductors or making more
efficient solar cells. In these materi-
als, the repeating pattern of the atoms
creates very complicated behavior
among the interdependent electrons.
And Babbush is figuring out how
quantum computers can be used to
make sense of these interactions.
If quantum computers succeed
in remaking our material world,
Babbush’s work will be one reason
why. —SIOBHAN ROBERTS
26 Change

server to be merged together in a master

AMAY model. It allowed the master model to train
BANDODKAR on data from every device without making it
AGE AFFILIATION
AG E A FFILIATION VIRGINIA Carnegie Mellon necessary to centralize that data. This not only
33
North Carolina
State University
SMITH 31 University reduced latency in the mobile experience but
INNOVATION
could also improve each user’s data privacy.
IN N OVATION
But combining millions of AI models also
His lightweight sensors Her AI techniques are efficient risks creating a central model that performs
could make wearable and accurate while preserving well on average but poorly for outliers—for
tech more useful and fairness and privacy. example, voice recognition software that fails
practical. when the speaker has an unfamiliar accent.
So Smith proposed a new technique for
more “personalized” federated learning.
Rather than merge a million localized models
Wearable technology can provide hen Virginia Smith began her PhD into one, it merges the most similar localized
real-time information about a per-
son’s health and fitness, but creat-
ing sensors that can collect data
without a cumbersome and imprac-
tical system of staying powered has
proved difficult. Amay Bandodkar
thinks he’s hit on a new way of cre-
ating “self-powered” biochemical
W in artificial intelligence, she had a
question: How do you train a neu-
ral network on data that is stored
across multiple machines?
Her attempts to answer it have made her a
leader in the field of federated learning, which
seeks to handle data spread across hundreds,
models into a few—the more heterogeneous
the data, the greater the number of final
models. Each model still learns from many
devices but is also tailored to specific subsets
of the user population.
Smith also works to overcome other
challenges in federated learning, such as
sensors through unconventional or even millions, of remote sources. accounting for different power and memory
technologies, making wearable Google researchers first introduced feder- constraints on different devices. To encourage
tech lighter and less cumbersome.
ated learning in 2017 to use with the compa- more research, she co-created an open-source
It’s about four times smaller and 20
ny’s mobile devices. The method they devised tool that lets researchers test their federated
times lighter than similar devices
produced two years ago, he says. involved training millions of neural networks techniques on more realistic data sets and in
The key to shrinking the sen- locally before sending them to a company more realistic environments. —KAREN HAO
sor was overhauling how it’s pow-
ered. “All the groups that were
working on this were using these
really bulky batteries, and the sen-
sor was around 3% of the total
size and weight,” he says. So he
built a sensor that doesn’t require
a battery: it harnesses the catalytic
properties of enzymes to gener-
ate signals without the need for
power sources. While this concept
can be used to develop self-pow-
ered sensors for some chemicals,
for other kinds of sensors that still
need a power source, Bandodkar
has developed a lightweight bat-
JESSICA JOHNSON (BANDODKAR); ILLUSTRATION SOURCE: COURTESY PHOTO

tery that runs on sweat. It’s made

of a magnesium anode and a cath-
ode made of silver and silver chlo-
ride, separated by a dry cellulose
membrane. When a person wearing
it starts to perspire, the cellulose
membrane absorbs the sweat and
acts as an electrolyte, effectively
turning the battery on and powering
the sensor.
Bandodkar has successfully
tested out a heart-rate sensor run-
ning on his battery, opening the
gateway for heart-monitoring wear-
ables. —NEEL V. PATEL
 35 Innovators Under 35 27

JIE XU
A GE AFFILIATIO N
Argonne
33 National Laboratory
I NNOVATION

She makes durable,

easy-to-manufacture
polymer semiconductors
for skin-like electronics.

Jie Xu has made printable, stretch-

able electronics viable for mass pro-
duction. Her multiple breakthroughs
could be used in future wearable
technology, advanced robotics, and
human-computer interfaces with
sensors connected to the skin.
The key for Xu was inventing
polymer circuits that kept working
despite being flexed, stretched,
and repeatedly moved. That had
been a challenge for researchers
until 2016, when Xu engineered a
two-polymer coating applied to
a rubbery surface that could be
stretched to twice its size and still
conduct electricity.
In 2019, she refined the tech-
nology so that her stretch- or weeks of computations using hundreds
able semiconductors could be
of specialized chips.
mass-produced using roll-to-roll
AGE AFFILIATION Xiao Sun is part of a research group at
manufacturing, a common indus-
IBM that has been finding ways to perform
trial fabrication process used to
print anything from textiles to plas-
XIAO SUN 34 IBM
those computations using three-digit, or
tics on large rollers. It was the first INNOVATION even just two-digit, numbers (in contrast, a
time anyone had achieved such a modern laptop or cell phone uses 20 digits
feat at scale.
He designs imprecise—but
energy-efficient—AI hardware to make calculations, while most dedicated
In the short term, Xu’s materi-
and software. machine-learning chips use five).
als and manufacturing inventions
The real trick is in finding techniques that
can make flexible displays and
skin-worn medical sensors much allow for small numbers to be used through-
more practical and easy to make. out the computation. You might still have to
Samsung Electronics has already do many trillions of computations, but each
patented two methods Xu helped one will be far simpler. This saves both time
define during collaboration with the rtificial-intelligence systems often and energy—using two-digit numbers is more

A
WES AGRESTA (XU); ILLUSTRATION SOURCE: COURTESY PHOTO

company. Xu’s materials could also

aid in the design of prosthetics with
functional skin-like outer coverings.
Wary of adding yet more plas-
tics into the world, Xu is search-
ing for versions of the polymer
semiconducting materials that
are recyclable or biodegradable.
“I think that kind of idea should be
integrated from the very beginning
of any commercial material,” she
says. —RUSS JUSKALIAN
require a vast amount of compu-
tation. That’s why in recent years,
AI hardware researchers have been
striving to achieve lower precision,
which is good enough to produce a correct
answer but avoids the use of calculations
that require keeping track of lots of digits.
Deep learning relies on networks that
might have dozens of layers, and millions,
or even billions, of parameters that must be
adjusted to the correct values, a process called
training the network. This often takes days
than 20 times more energy efficient than
doing the same calculations using numbers
in the billions, according to a paper by Sun
and colleagues at IBM.
In February, IBM announced a new chip,
based in part on Sun’s work, that trains neu-
ral networks using computations involving
mostly three-digit numbers. The company
hopes to use it not only to train large neural
networks in cloud computing centers but also
in mobile phones that could train on local
data. —PATRICK HOWELL O’NEILL
28
AGE AF F I L I A T I O N
YICHEN
SHEN 32 Lightelligence
I N NOVATION
Optical chips that can make
calculations for neural networks are
poised to become big business.
here are two basic types of computations
T involving neural networks. First, the networks
must be trained, which usually involves show-
ing them lots of data, causing them to adjust
the strength of the connections between their
numerous “neurons.” Next, those existing connec-
tions are used to make decisions. It’s the difference
between learning to drive and driving.
The difference is crucial. If a neural network takes
weeks to learn how to recognize images, that’s not
necessarily a problem. But if it is driving an autono-
mous car, it needs to be able to make life-or-death
inferences in fractions of a second.
That’s where optical computers come in. Despite
decades of research, they’ve never worked that well.
It’s harder to manipulate photons than electrons. But
for certain types of computations—like those com-
monly needed when using an existing neural net-
work to make inferences—photons are just the thing.
In 2017, Yichen Shen and Nicholas Harris (at right)
published a widely cited paper on the use of optical
circuits for machine-learning tasks including speech
and image recognition. Their design, one review arti-
cle notes, “represents a truly parallel implementation
of one of the most crucial building blocks of neural
networks using light, and modern foundries could
easily mass-fabricate this type of photonic system.”
This means that optical computers on a chip could
become a huge business, with one in every device that
uses a neural network to make decisions.
Shen and Harris now run competing startups.
Shen’s firm, Lightelligence, released a prototype optical
AI chip in 2019, and Shen says they have secured over
$100 million in funding. —KONSTANTIN KAKAES
Change
AGE AFFILIATION
NICHOLAS
HARRIS 33 Lightmatter
INNOVATION
Shining light through optical chips
might be the fastest way for neural
networks to make decisions.
or decades physicists and engineers have
F dreamed of making optical chips that use pho-
tons, not electrons, to do computing. Such cir-
cuits could be lightning fast and energy efficient.
But making them work has been difficult.
In 2017, Nicholas Harris, together with Yichen
Shen (at left) and other colleagues at MIT, published
a widely cited paper describing a design that allowed
them to calculate the outputs of neural networks that
had been conventionally trained.
The paper describes a circuit of 56 programma-
ble interferometers—devices that carefully break
apart and recombine light waves. The circuits they
created solved a simplified problem of recognizing
vowels correctly—distinguishing about three quarters
of the 180 cases they tried. This wasn’t as good as a
conventional computer, which got over 90% of them
right. Shortly thereafter, Shen and Harris launched
competing startups.
Once a given neural network has been trained
and implemented on an optical chip, performing
inferences— figuring out which vowel corresponds
to which sound, or how an autonomous car should
react if a pedestrian steps into the street—can be
COURTESY PHOTO (SHEN); BOB O’CONNOR (HARRIS)
almost as simple as shining light through it. This has
the advantage of being both fast and energy efficient.
In March 2021, Lightmatter announced it would
soon start selling a “machine-learning accelerator”
chip. “It’s just a completely different kind of com-
puter,” says Harris. “Right now we’re at about a factor
of 20 times more efficient than the most advanced
node in digital computers.” Lightmatter closed a
second round of funding in May, bringing its total
investment to $113 million. —KONSTANTIN KAKAES
 35 Innovators Under 35 29

work with, immunotherapy drugs fail against

JONATHAN these tumors.
GOOTENBERG As a graduate student at Stanford,
AGE AFFILIATION
A GE AFFILIATIO N SHELLEY Bolt Ackerman worked with Edgar Engleman,
30 MIT
ACKERMAN 29 Biotherapeutics a professor of medicine and pathology, to
INNOVATION
develop a therapy aimed at turning cold
I NNOVATION
tumors into hot ones. The approach uses a
Expanding the capabili- She co-invented a novel tumor-targeting antibody chemically attached
ties of gene editing. immunotherapy for to an immune-stimulating small-molecule
difficult-to-treat cancers. drug that prompts the immune system to
recognize and attack the tumor, transform-
ing it into a hot one invaded by tumor-kill-
ing T cells. Engleman founded a biotech
company, Bolt Biotherapeutics, in 2015 to
The gene-editing tool CRISPR sing the body’s own immune system commercialize the approach; Ackerman
uses a protein called Cas9 to
snip out a targeted part of the
genome. It does amazing work but
has downsides. It can cause unin-
tended edits to other places in the
genome, and if you want to make
only a temporary tweak, Cas9 can’t
do it.
U to fight cancer has shown promise
against several types of tumors, but
it’s not always effective. “There’s a
whole subset of patients that it really
doesn’t work well in,” says Shelley Ackerman.
Tumors have to be “hot,” or inflamed,
for immunotherapy drugs to work well.
joined Bolt in 2018.
As a child, Ackerman lost her uncle and
a close friend to metastatic cancer within a
year, and that experience made her want to
keep working on the therapy in hopes that
it would one day be used to treat patients.
Last year, Bolt began testing its approach
Jonathan Gootenberg is cre- Hot tumors are characterized by the pres- in patients with breast, gastric, and other
ating other editing tools to get ence of a type of immune cell called T cells. tumors that express a protein known as
around these shortcomings and
Immunotherapy drugs give those T cells a HER2. The company, which has raised $438
add to the capabilities of CRISPR.
boost, making them better cancer fighters. million in funding, is also developing drugs
Gootenberg has used Cas12—a
more compact protein than Cas9— But many tumors are “cold” and thus evade for colorectal, lung, and pancreatic cancers.
to edit many genes at a time. This the immune system. Without any T cells to —EMILY MULLIN
capability could be used to edit a
patient’s immune cells so that they
fight cancer.
Then there’s Cas13. Gootenberg
and his colleague Omar Abudayyeh
(a 2020 Innovator Under 35)
demonstrated that the protein could
target RNA instead of DNA—an
intriguing finding. Many viruses
have RNA as their genetic material,
and bacteria have both DNA and
RNA, so the researchers reasoned
that you could use Cas13 to find
genetic material from pathogens
in human cells. That led to repur-
posing the gene-editing tool as a
paper-based diagnostic test, and in
2019 Gootenberg and Abudayyeh
cofounded Sherlock Biosciences to
JUSTIN KNIGHT (GOOTENBERG); JORDAN ENGLE (ACKERMAN)

commercialize the technology.

—EMILY MULLIN
 30
Change

GUTTER CREDIT HERE

 35 Innovators Under 35
Their creative
approaches to
innovation are
making the world
a more equitable
place.
EMMA PIERSON
AGE AFFILIATION
30 Cornell University
IN N O V A T I O N
She employs AI to get to
the roots of health
disparities across race,
gender, and class.
Cornell University computer sci-
entist Emma Pierson uses AI and
emerging data science models to
reveal how health disparities arise
between sexes, races, socioeconomic
groups, and other demographic cat-
egories. “These are fancy ways of
saying I use math to find patterns
in large data sets, and the specific
types of patterns I’m looking for
are attempting to answer sort of
old questions in health and social
sciences,” she says.
The “old questions” she’s inves-
tigating range widely in their specif-
BOB O’CONNOR
ics, but she focuses on uncovering
how systemic inequalities in public
health come to be, and pointing at
ways to dismantle them. For exam-
ple, by analyzing mobile-phone data,
she recently showed that particu-
lar “superspreader” locations were
primarily responsible for transmit-
ting covid-19 across populations,
and that low-income and minority
communities suffered greater risk
of exposure.
Beyond the pandemic, Pierson’s
research team recently examined
nearly a decade’s worth of data to
show the extent of racial dispari-
ties in traffic stops made by police
across the US. She analyzed men-
strual health data from millions of
women in 109 countries to demon-
strate how effects on mood and
behavior are experienced univer-
sally, seeking to destigmatize dis-
cussions around women’s health.
And she used deep learning to study
data on knee pain, revealing that the
problem was often poorly measured
and even exacerbated in patients
from racially underserved groups
and lower economic backgrounds.
Pierson has made it her mission
to see this work break out of the
31
confines of academia. She’s a fairly
regular contributor to the New York
Times and the Atlantic, offering up
a layperson’s account of her work to
a large audience. And she engages
directly with organizations that can
pressure policymakers. Her work on
racial disparities in traffic stops ulti-
mately led the Los Angeles Police
Department to announce that it
would reduce the number of ran-
dom stops it conducted, and state
departments of health leaned on
her covid-19 findings to determine
how to safely reopen businesses.
A self-professed math nerd,
Pierson simultaneously earned a
bachelor’s degree in physics and
a master’s in computer science at
Stanford before moving to Oxford as
a Rhodes scholar, where she earned
a master’s in statistics, afterward
earning a doctorate in computer
science at Stanford.
“I wanted to work on problems
that were very concretely tied to
people’s lives,” she says. “I think
this sense was particularly driven
by my own family’s medical history.”
In December 2011, Pierson learned
she was carrying a genetic mutation
that increases her risk of breast and
ovarian cancer, and it drove her to
focus on work that could make an
impact in health care and medicine.
Industries like health care deal
with insanely large data sets that
can really be understood only with
the kinds of analytical techniques
Pierson has mastered. The data
might be the genomes of thou-
sands of people, containing mil-
lions upon millions of data points,
or medical images from many dif-
ferent patients, representing tera-
bytes of information. AI tools can
sort through this data and look for
patterns that no human could readily
identify. “Computational methods
are not optional here,” Pierson says.
“They’re the only solution.”
—NEEL V. PATEL
32
AGE AFFILIATION
SHRIYA
SRINIVASAN 27 MIT
I N NOVATION
Her surgical techniques provide
a sense of touch to people with
prosthetic limbs.
s a child, Shriya Srinivasan, now a
A postdoctoral researcher in biomed-
ical engineering at MIT, witnessed
the challenges of living with pros-
thetic limbs. A friend had been born
with missing limbs and used prosthetics; like
amputees, whose nerves have been severed,
her brain lacked the important neural sig-
nals that enable most people to feel objects,
maintain balance, and sense their body’s
position in space. Srinivasan has invented
two new types of surgical techniques that
could soon help people using prosthetic
limbs regain their sense of touch.
Change
Her first innovation, which she developed
as an MIT doctoral student, involves grafting
small segments of muscle onto the resid-
ual limb; the goal is to enhance the mind’s
awareness of limb position and movement.
Patients who underwent a version of this
procedure in clinical trials have exhibited
far greater control over their prostheses—
and less pain—than those with traditional
amputations. Her second procedure has
shown early promise in re-creating touch.
It works by fitting a person’s residual limb
with flaps of skin from the fingertips or feet,
encased by a muscle graft and electrode. A
prosthetic arm or hand is then equipped with
sensors and a wireless transmitter; when
it touches an object, it conveys that sensa-
tion to the natural sensors on the grafted
skin—which relay it on to the brain. Both
techniques can be performed either as part
of an amputation procedure or in patients
with previous amputations.
Ultimately, Srinivasan hopes her work
will help make using a prosthetic limb far
more like the real thing—while initiating a
broader shift in our approach to amputation
from a form of salvage to a method of restor-
ing mobility. —JONATHAN W. ROSEN
SRIRAM
CHANDRASEKARAN
AGE AFFILIATION
34 University of Michigan
INNOVATION
His AI
systems
identify better
treatments for
tuberculosis.
Before covid-19, tuberculosis was
the most dangerous infection in
the world, killing more than 1.5 mil-
lion people annually. The problem
prompted Sriram Chandrasekaran
to build AI tools to identify potent
drug combinations to treat it. His
goal is to boost the effectiveness of
existing antibiotics to combat drug
resistance among TB patients.
Drug-resistant infections occur
when people don’t finish their
course of treatment or are treated
incorrectly. They can also occur
when people come in contact with
a patient infected with drug-resis-
tant bacteria. While a typical TB
treatment regimen lasts six to nine
months, a drug-resistant case takes
18 to 24 months to treat. Chan-
drasekaran wants to drastically
reduce this timeline. Curing patients
faster could also save thousands of
ILLUSTRATION SOURCE: JIMMY DAY (SRINIVASAN); AKHIL KANTIPULY / MICHIGAN ENGINEERING (CHANDRASEKARAN)
dollars in treatment costs.
Chandrasekaran’s systems pre-
dict the effectiveness of various
drug combinations for TB. “We’ve
found some really surprising ones,”
he says, including an antipsy-
chotic drug that would enhance the
potency of existing antibiotics. He
and his team confirmed the results
against the TB bacterium in the lab.
Many drugs work in the lab but
aren’t effective in the body, and
Chandrasekaran wanted to make
sure his algorithms take this into
account. One system he built sim-
ulates characteristics of the infec-
tion site—for example, how much
oxygen it gets or whether amino
acids are present, which can affect
a drug’s effectiveness. Chandrase-
karan’s lab is now identifying prom-
ising drug combinations for use in
clinical trials of treatment against
drug-resistant TB. —EMILY MULLIN
 33

AADEEL
AKHTAR
AGE AFFILIATION

34 Psyonic
INNOVATION

His bionic hands

combine sensitivity
with affordability.

Aadeel Akhtar has developed algo-

rithms that make upper-limb pros-
thetics much more functional to use.
Some send electrical currents to
stimulate the nerves so that users
can “feel” what their prosthetics are
photo and calligraphy TK touching; others record the elec-
trical currents caused by muscle
contractions, making it possible to
control movement. Akhtar has been
doing this work for over 10 years,
first as a doctoral researcher at
the University of Illinois at Urbana-
Champaign and then, starting in
2015, as the founder of the robotic-
limb startup Psyonic.
Akhtar holds four patents on
advances in prosthetics that have
all gone into Psyonic’s first prod-
uct, the Ability Hand. The Ability
Hand was designed to be con-
trolled by both muscle sensors and
Bluetooth (yes, there’s an app!)
and provide tactile sensory data
to its user, all while withstanding
the normal stresses of everyday
life (like getting knocked against a
table) without cracking.
Akhtar’s team of 20 designed
with affordability in mind, he says,
and built a hand inexpensive
enough to be covered by Medi-
care. This means far more peo-
ple in the US will be able to afford
it. Previously, Akhtar explains,
the only insurance that covered
bionic hands was associated with
GLEN GYSSLER (DEVICE); SKOT WIEDMANN (AKHTAR)

veterans’ benefits and worker’s

compensation claims, which he
estimates cover only about 10%
of the need in the United States.
Participation from Medicare would
make bionic hands available to
75% of individuals in the US who
need them. “If Medicare covers it,
then other insurers usually follow
suit,” Akhtar says. —EILEEN GUO
34
They’re looking
to the future
of quantum
computing, energy
policy, robotics,
and more.
KAYLA LEE
AG E AFFILIATION
30 IBM
I NNOVATION
She’s working to build a more
diverse future for quantum
computing.
n 2018 Kayla Lee joined the enterprise
I consulting group at IBM, where part
of her job is to persuade clients they
should be interested in quantum com-
puting. For each client, she says, she
needs to figure out the same thing: “How
do you make this new technology that is a
little bit complicated, and sounds kind of
like a science project, relevant to them?”
There are parallels between that work
and her other project: leading the launch
of the IBM-HBCU Quantum Center, a
partnership between the company and 23
historically black colleges or universities,
which aims to make quantum computing
more accessible to Black students and
faculty. Lee wants to give Black STEM
students and scholars the foundation to
excel in this emerging field.
Change
Through the partnership, HBCUs have
access to IBM’s cloud-based quantum
computing service, which undergraduates,
graduates, and faculty can use for research.
The partnership not only supports Black
faculty working on quantum projects but
provides funding to “seed these research
projects,” says Lee. In one example, IBM
recently partnered with the International
Society for Optics and Photonics to cre-
ate a faculty award in quantum optics and
photonics specifically for IBM-HBCU
Quantum Center members.
Lee sees the project as a way to sup-
port Black students in an area where
they’re grossly underrepresented. In 2017,
Black students were awarded just 3% of
all bachelor’s degrees in physics in the
United States, and only 2% of physics
PhDs. What’s more, according to the
National Science Foundation, a third of opportunities for engagement” and sim-
all Black students who have earned doc- ply show students “what quantum scien-
toral degrees got their bachelor’s degrees tists look like.” This is especially important,
at HBCUs, but to date few HBCUs have she says, because quantum computing is
offered opportunities for students to such a young field. “We really are at the
PETER GARRITANO
study or conduct research in quantum start of a new model of computation, in
information. the same way that we were at the start of
Lee aims to change that. She wants a new model ... back in the ’60s,” she says.
the Quantum Center to create “clear “So the questions we’re asking today are:

What do the qubit implementations look
like? How do we make less noisy qubits?
What does that architecture look like?”
But for Lee there’s a further question
about quantum computers: “I’m more
COURTESY PHOTO
focused on who gets to use them.”
The question of who gets the oppor-
tunities to work on this cutting-edge
technology will shape the way the field
35 Innovators Under 35
develops. She points to artificial intel-
ligence, which is already known to be
afflicted by problems with racial bias.
She says this problem could be exponen-
tially worse in quantum computing, both
because of the complexity and inscruta-
bility of the machines and because “there
are even fewer representative people” in
the field. —EILEEN GUO
35
VARUN
SIVARAM
AGE AFFILIATION
Biden-Harris
32 administration
INNOVATION
Designing new public
policies to promote
energy innovation.
Varun Sivaram earned his doctorate
researching novel solar materials, but
when he graduated in 2013, it wasn’t
clear where he could apply those
skills in the private sector.
Very few startups working on
advanced approaches had survived
the clean-tech bust of the early
2010s. Commodity silicon solar pan-
els, mostly made in China, dominated
the business.
That experience prompted him to
begin exploring what changes to the
innovation system would be required
in order to develop better and cheaper
clean energy technologies. In stud-
ies and books, Sivaram argued that
governments must provide far more
funding and early policy support for
crucial technologies. He also con-
cluded that solar power would still
require significant advances to gener-
ate an ever larger share of electricity.
He worked on these issues
directly as chief technology offi-
cer at ReNew Power, a large Indian
renewable energy company. Now
he’s joined the Biden administration
as a senior advisor for energy inno-
vation to John Kerry, the US climate
czar. Sivaram traveled to India with
Kerry, who negotiated a partnership
to help that nation achieve its 2030
climate goals. Those include reaching
450 gigawatts of renewable capacity.
Sivaram believes that innovation
is the most powerful lever the US has
to help the rest of the world raise its
climate ambitions. Driving down the
cost of carbon-free technologies
makes it cheaper, easier, and more
politically palatable to accelerate the
shift to emissions-free energy. Sivaram
adds that this is particularly crucial for
poorer nations, which often can’t afford
to sacrifice economic growth. Without
such advances, emissions in emerging
economies will soar in coming decades,
he warns. —JAMES TEMPLE

EMMA BEEDE
AG E A FFILIATION
30 Google
IN N OVATION
Her work helps
ensure that fancy
AI tools perform
in the real world.
Emma Beede has an unorthodox
claim to technological fame: a
study she ran showed that one
of her employer’s new technol-
ogies was unfit for actual use in
the real world.
Beede’s study tested a
deep-learning algorithm created
by Google Health to screen eye
images for diabetic retinopathy,
a condition caused by high blood
sugar that damages the retina and
SARA BERGER
AG E AFFILIATION
33 IBM Research
I NNOVATION
Employing machine
learning to make
pain management
more accessible.
Developing smart technology to
help patients assess and manage
pain is a deeply personal pursuit
for Sara Berger, who spent years
watching her parents cope with
chronic pain and struggle to nav-
igate the medical system. “A lot of
the suffering from having chronic
pain is about no longer having con-
trol over your body and your body’s
sensations,” says Berger. “Being
able to use digital technologies
provides a sense of control and
creates more informed conver-
sations with physicians.”
A neuroscientist at IBM’s T.J.
Watson Research Center, Berger
makes it difficult to sense light.
Beede found that the algorithm,
which had performed with over
90% accuracy in the lab, bombed
in real-world tests across 11 clinics
in Thailand. She found that this was
because the algorithm was trained
on high-quality eye scans, and
when the quality of images taken
in the clinic suffered because of
factors like poor lighting, the scans
were rendered useless. More than
20% of retinal scans were rejected,
leaving frustrated patients and
their health-care providers looking
for more conventional alternatives.
Beede thinks such unsatis-
fying results are a critical exam-
ple of the need to ensure that
AI-powered tools for humans
are put through rigorous and
meticulous testing before being
deployed. “Humans in the real
world are complicated, and we
should account for that,” she says.
“We need to be doing our due dili-
gence to study those downstream
effects so that we can mitigate
any risk for harm.”
—NEEL V. PATEL
employs machine learning to quan-
tify long-term pain and help predict
ways to relieve it. With wearables
and environmental sensors, she
can capture metrics including heart
rate, sleep patterns, and even the
acoustic properties of a patient’s
speech, all of which provide data
about the person’s pain experience.
Those metrics can then be analyzed
using machine learning, taking into
consideration other factors such as
the emotional toll that often results
from chronic discomfort, decreased
mobility, or lost time with loved ones.
What results is a far more holistic
and informed assessment and treat-
ment plan than those informed by
traditional pain scales, which are
prone to bias and oversimplifica-
tion. “Pain isn’t linear,” says Berger.
“Our assessment of it shouldn’t
be either.”
Many people with chronic con-
ditions, especially women and peo-
ple of color, feel marginalized by
the health care system and expe-
rience bias when they seek treat-
ment for pain. “I’m on a mission to
transform pain management into
an accessible, personalized, and
trusted experience for individuals
across different socioeconomic
backgrounds,” says Berger.  he should move on to some-
—KATHRYN MILES
PRIYA DONTI
AGE AFFILIATION
Carnegie
28 Mellon University
INNOVATION
Finding climate-
change solutions
via computer science
and public policy.
Priya Donti knows that a problem
as complex and pervasive as cli-
mate change won’t be solved by
one discipline alone. That’s why
she cofounded Climate Change
AI, an interdisciplinary organiza-
tion that brings together aca-
demics and industry experts to
demonstrate how machine learn-
ing can help.
Donti’s work combines com-
puter science, engineering, and
AÄRON VAN DEN OORD
AGE AFFILIATION
33 DeepMind
INNOVATION
His AI system
creates artificial
voices that sound
remarkably human.
In 2016, Aäron van den Oord
had just won an award for his
research in image generation
when he was struck by an idea.
If his technique could learn
to predict a two-dimensional
sequence of pixels, could it
also learn to predict a wave-
form and thus generate realistic
voices? The idea was intriguing
but seemed like a long shot.
His manager at DeepMind, an
AI research subsidiary of Goo-
gle, gave him two weeks to try it
out, saying that if it didn’t work,
thing else.
public policy, and her research
focuses on how electric grids can
more reliably integrate renew-
able energy.
In 2019, Donti was also a lead
author on an influential paper
titled “Tackling Climate Change
with Machine Learning.”
“The tremendous response we
received from that paper demon-
strated just how many people
felt a moral obligation to work
on climate change but who also
felt like they lacked the neces-
sary community to do that work,”
says Donti.
Donti, a second-generation
Indian American says she’s well
aware of the immense burden
already felt by some of the plan-
et’s most vulnerable people and
recognizes that climate change
will only exacerbate those bur-
dens. “We know that the world’s
most disadvantaged populations
are going to be disproportionately
affected by climate change,” says
Donti. “Climate Change AI wants
to help mitigate that.”
—KATHRYN MILES
The results beat everyone’s
expectations. Within two weeks,
van den Oord had a prototype.
Within three months, it was gen-
erating more realistic voices than
any existing systems. Within
another year, Google had begun
using WaveNet, as the system
came to be called, to generate
voices for Google Assistant.
WaveNet now powers 51 voices
as well as Google’s newest voice
assistant, which calls salons and
restaurants on behalf of users to
book appointments or reserve
tables. The results are startlingly
realistic. When Google CEO Sundar
Pichai first demoed Duplex in 2018,
with all its human-like “umms” and
“ahs,” it set a new bar for what can
be possible when people commu-
nicate with machines.
While voice assistants need to
do more than just generate a syn-
thetic voice—they also need to be
able to recognize when someone
KRELL INSTITUTE (DONTI); COURTESY PHOTOS
is talking and understand what’s
being said, each of which is a
challenge unto itself—research-
ers have long sought to create the
right artificial voice for achieving
natural and engaging conversa-
tions. “There’s a lot of meaning
in a voice,” says van den Oord.
—KAREN HAO

AGE AFFILIATION
KAITLYN National Institutes
SADTLER 31 of Health
INNOVATION
Her test was among the first to
determine how many people had
been infected with covid-19.
In early 2020, Kaitlyn Sadtler envisioned
a long, slow season getting her lab up and
running. Then covid-19 happened. Within
weeks, she and her team were among the
first to develop an effective antibody assay
DORSA SADIGH
A GE AFFILIATION
NIH/CHIA-CHI CHARLIE CHANG (SADTLER); SINEAD DUBEAU (ELLIS); HECTOR GARCIA-MOLINA (SADIGH)
30 Stanford University
I NNOVATION
She uses simulated environ-
ments to teach robots to be
better collaborators with people.
By developing new ways for computers to antic-
ipate people’s actions, Dorsa Sadigh wants to
help pave the way for a future in which human
and robots do things like share the roads.
In one widely cited paper from 2016, she
and her colleagues considered the idealized
case of two cars, one driven by a person and
another by a computer program. She first had
real people drive a car in a video-game-like
simulation with several autonomous coun-
terparts that followed preplanned routes. On
the basis of people’s behavior in the simula-
tion, she developed a model for how humans
drive, which the robot driver then used to
35 Innovators Under 35
capable of determining how many people had
been infected with covid-19, whether they’d
shown any symptoms or not.
Antibodies tag viruses for destruction and
help the body mount an immune response.
Those antibodies can linger for months.
Existing tests didn’t pinpoint the unique anti-
bodies for the covid-19 virus, leading to false
positives among people who had previously
been exposed to other coronaviruses. Sadtler
and her team at NIH made a highly sensitive
antibody test, which uses six different assays
to more accurately identify the presence of
covid-19 antibodies. Early results published
in January confirmed that about 16.8 million
Americans had been infected with covid-19
but hadn’t been diagnosed. (Sadtler will update
those findings this fall and estimates that as
many as one-third of all Americans have been
infected with the virus.)
The blood test is sensitive enough to
determine whether an individual has anti-
bodies from the virus itself or in response
to a vaccine, and it can distinguish between
variants of the virus as well. It’s simple and
cheap to use, making it practical in both
rich and poor countries. “This is a global
pandemic,” says Sadtler, “which means we
need to think globally.” —KATHRYN MILES
devise new strategies for interacting with
them. Without ever being explicitly told to
do so, it did things like slowly backing up at
an intersection, encouraging the “human” to
go first. It also developed an attitude, learning
how to cut human drivers off or force them
to change lanes by swerving toward them.
More recently Sadigh and Dylan Losey,
at the time her postdoctoral student, taught
robots in a simulated setting how to trick
humans in a game that involves negotiating
who will do more work in carrying plates to a
table. “This robot is capable of bringing two
plates, but misleads the human to believe that
it can only carry one in order to reduce its
overall effort,” they wrote in a paper on the
work. Teaching robots to be lazy might not
sound particularly worthwhile. But Sadigh
and Losey are thinking of future applications
in which robots might be called upon to help
stroke patients in their recovery, for example.
Robots, they say, “need to make intelligent
decisions that motivate user participation.”
—WILL DOUGLAS HEAVEN
37
LEAH
ELLIS
AGE AFFILIATION
31 MIT, Sublime Systems
INNOVATION
A new, climate-friendly
way to make cement.
Making cement is one of the single
largest drivers of climate change,
accounting for almost a tenth of
global carbon dioxide emissions.
Ground-up limestone is typically
cooked together with sand, clay,
and other materials in kilns that are
heated to around 1,500 ˚C (2,700 ˚F).
The limestone releases carbon
dioxide as it breaks down, as do
the fossil fuels that are burned to
achieve those temperatures. For
every resulting pound of cement,
roughly a pound of carbon dioxide
escapes into the atmosphere.
Leah Ellis came up with a better
way. Sublime Systems, a startup
she cofounded in March 2020, dis-
solves pulverized limestone in water
and then applies an electric cur-
rent to trigger a series of chemical
reactions.
The general idea of using elec-
tricity rather than heat to break
down limestone has been around
for a while, though earlier attempts
worked at higher temperatures. Sub-
lime’s apparatus operates at room
temperature. Lots of carbon dioxide
is still released from the limestone,
but it’s much easier to capture and
reuse—the gas comes out of one
end of the device, mixed with oxy-
gen, while hydrogen gas is released
from the other end.
This electrochemical reaction
produces pure lime, a white powder
made of calcium, oxygen, and hydro-
gen. It can then be cleanly cooked in
a kiln with silicon and oxygen to make
cement. Ellis and her colleagues are
still considering a variety of potential
business models. Because they can
rely on increasingly cheap electricity
from solar or wind farms, Ellis says,
they’ll be able to match the prices of
standard cement. —JAMES TEMPLE
38
They’re making
crucial advances
in fusion power,
computing,
biosensors, and
robotics.
DAVID ROLNICK
AG E AFFILIATION
30 McGill University
I NNOVATION
He’s employing artificial
intelligence in the fight against
climate change.
In 2019, as a postdoctoral researcher at the
University of Pennsylvania, David Rolnick
was lead author of an influential report that
described various ways machine learning
could reduce greenhouse-gas emissions
and help society adapt to climate change,
from predicting energy needs to man-
aging forests to modeling planet-scale
weather systems. His coauthors included
DeepMind cofounder Demis Hassabis and
Turing Award winner Yoshua Bengio. That
year, Rolnick was a lead organizer for the
first workshops on climate change at three
leading AI conferences, and lead organizer
of an event on AI at the United Nations
Climate Change Conference.
“David Rolnick has been hugely influ-
ential in convening AI practitioners to
Change
work on climate change,” says Andrew
Ng, a cofounder of Google Brain and for-
mer chief scientist at Baidu. “By helping
shape a vision of how AI could help cli-
mate change and tirelessly organizing a
community around it, he has catalyzed
a significant amount of activity on this
important topic.”
Rolnick now leads a group at McGill
University that uses different AI techniques
to attack problems related to climate.
For example, data relevant to climate
change—records of infrastructure spend-
ing or greenhouse-gas emissions or sim-
ply weather patterns—varies enormously
between countries. And yet climate needs
to be understood at a global level.
“In the Global South there can be
less information on infrastructure,” says
Rolnick. “So policymakers may have less to
go on when it comes to making decisions
about energy requirements or managing
coastal flood risk.” Countries also have
different regulations about what does and
does not get recorded. Germany gathers
information on where its solar panels
are, for example, but the US does not, so
researchers are using machine learning
to identify solar panels in the US from
satellite imagery. Machine learning can
also be used to forecast energy demand
more accurately than is possible with
existing techniques, Rolnick says. This
allows energy providers to manage their
electricity grids more efficiently.
Rolnick and his colleagues are trying
to come up with new machine-learning
techniques that could be applied to the
study of climate change as well.
For instance, they are building algo-
rithms for transfer learning, which involves
training an AI on one set of examples
and then transferring what it’s learned to
new situations. They are also researching
meta-learning, a set of techniques that
make AI better at learning from small or
incomplete data sets. Rolnick thinks these
methods are especially useful for modeling
biodiversity because sources of real-world
data are so patchy.
Rolnick is also involved in projects that
combine machine learning with climate
models to simulate complex physical and
atmospheric processes like cloud forma-
tion. The precise means by which clouds
form, and how much they reflect or absorb
sunlight, is one of the largest sources of
uncertainty in existing climate models—
partly because simulating clouds in cli-
mate models is computationally intensive.
Using machine learning to find patterns
in when and where clouds form and how
reflective they tend to be—without trying
to understand the underlying atmospheric
chemistry—allows scientists to run mod-
els more quickly.
Rolnick and his collaborators are con-
vinced AI will be a crucial tool in fighting
climate change. All the same, there are
growing concerns that machine learning
itself is part of the problem. He acknowl-
edges that training today’s largest AI models
consumes large amounts of energy, but he
points out that this contributes a tiny frac-
tion of global emissions—and that the real
climate risks from AI arguably have more to
do with its uses in areas such as oil and gas
GUILLAUME SIMONEAU
exploration. “I’m much more worried about
negative applications of machine learning
than I am about its energy use,” he says.
—WILL DOUGLAS HEAVEN
 35 Innovators Under 35 39

GUTTER CREDIT HERE

40 Change

Modern chips are composed of millions

MOSES or even billions of components. Some per-
NAMARA form computations; others store data in
AGE AFFILIATION
AG E A FFILIATION ANNA Google Brain / short-term memory. Figuring out the best
29 Clemson University
GOLDIE 27 Stanford University way to place all the components in a chip’s
INNOVATION
layout can take engineers weeks or even
IN N OVATION
months—they must try to minimize power
Working to break down She uses AI to design consumption and area but also maximize
the barriers keeping microchips much more performance, all while making sure that
young Black people from quickly than humans can. traffic between components doesn’t get
careers in AI. too congested.
Goldie’s AI can, in under six hours, come
up with solutions that match—or even out-
perform—the ones that people were able
Moses Namara knew two funda- nna Goldie designs computer chips to develop.
mental truths: first, that misuses of
AI disproportionately harm Black
communities around the world,
and second, that Black people are
underrepresented in university AI
programs. Just 1.8% of students
enrolled in computer science PhD
programs in the United States
A using reinforcement learning, an AI
technique that works by repeatedly
generating solutions from an artificial
neural network. The system then pro-
vides feedback to the network, “reinforcing”
pathways that lead to successful outcomes
and weakening pathways that don’t.
In early 2021, Goldie collaborated with
Google engineers to produce physical
versions of her layouts for Google’s latest
artificial-intelligence chip. By using AI to
design better hardware faster, she hopes to
pave the way for AI advances that further
improve and accelerate hardware design,
were Black in the 2018-2019 Building on this branch of machine learn- creating a symbiotic loop between hardware
school year, and the numbers were ing, which also underlies the most success- and artificial intelligence.
only marginally better for master’s
ful methods for teaching computers to play “It generates these very strange,
students.
games like chess or Go, has allowed Goldie alien-looking layouts,” she says. “The chip
Namara knew something else,
too: that the barriers to entry are and her team to speed up the process of designers were like: What if it goes wrong?”
often rooted in resources, and that chip design. It didn’t. —WILL DOUGLAS HEAVEN
some of those resources were
things a mentorship network could
provide. “One is just information,”
he says. For example: applicants
need to know which research
opportunities to pursue as under-
grads, which university programs
and professors best suit their inter-
ests, and what resources might be
out there to help with the expen-
sive process of actually applying.
“If you don’t know where to look
for the information, then that’s the
number one step that you’re going
to fail,” he says.
So in 2018 Namara co-created
Black in Artificial Intelligence to
help students applying to gradu-
ate school. Black in AI has men-
tored 400 applicants, 200 of
COURTESY PHOTO (NAMARA); PRIYANKA SHARMA (GOLDIE)

whom have been accepted to

competitive AI programs. It pro-
vides an array of resources: men-
torship from current PhD students
and professors, CV evaluations,
and advice on where to apply.
Namara now sees the mentorship
system evolving to the next logical
step: helping Black PhD and mas-
ter’s students find that first job.
—ABBY OHLHEISER
 35 Innovators Under 35 41

GEORGE
BOATENG
A GE AFFILIATIO N

28 SuaCode.ai
I NNOVATION

He built a smartphone-
based platform to teach
young people to code—
and tackle Africa’s IT
skills gap in the process.

George Boateng’s venture,

SuaCode.ai, emerged largely by
accident. In 2013, as an under-
graduate at Dartmouth College,
he’d teamed up with a group of
friends to launch a summer inno-
vation boot camp for high school
students in their native Ghana.
When the donated laptops they’d
gotten for the course broke down
a few years later, they were in a fix:
only a quarter of the students had
laptops of their own, and buying
more would overwhelm their bud-
get. All the students, however, had
smartphones—so Boateng and his
colleagues redesigned the coding
module to fit a five-inch screen.
The experience went so well chips. But turning carbon-nanotube transis-
that it hatched a spinoff: in 2018,
tors and wires into actual devices has proved
Boateng and cofounder Vic-
AGE AFFILIATION difficult, and Shulaker has solved several
tor Kumbol ran their first pilot MAX problems to make them work. He devel-
of SuaCode, an eight-week
smartphone-based course. The
SHULAKER 33 MIT
oped a way to remove poorly formed carbon
course, which teaches Processing, INNOVATION nanotubes during production, devised new
a Java-based language, now has processes to create wafers of nanotube-based
more than 600 graduates from two
His work with carbon nanotubes
could lead to the next generation transistors using regular industrial fabrication
dozen countries. Boateng, currently
of computers. plants, and invented a new design ensur-
a doctoral candidate in applied
ing that chips built with a certain number
machine learning at ETH Zurich,
of defective tubes are guaranteed to work.
RAENG/GGIMAGES/FRANCISKOKOROKO (BOATENG); COURTESY PHOTO (SHULAKER)

also engineered an English- and

French-speaking AI-powered teach-
ing assistant named Kwame—a nod
to Ghana’s first president, Kwame
Nkrumah. “His pan-Africanist vision
resonates with our goal of empow-
ering youth across the continent,”
Boateng says.
Boateng’s hope is that the
automated nature of the course
will help it reach far more stu-
dents—providing early exposure
to coding that will serve as a bed-
rock for further education and ulti-
mately help them land well-paying
jobs in tech.
—JONATHAN W. ROSEN
Max Shulaker has built the world’s first func-
tional computer using carbon nanotubes, and
he has also designed systems that combine
computing, memory, and sensing directly on
top of one another on a single chip. Together,
these new technologies could increase energy
efficiency in computers up to 1,000-fold and
make possible a whole world of new devices
like low-cost medical sensors.
Carbon nanotubes are “basically a straw
that is one carbon atom thin,” says Shulaker.
For 20 years, researchers have talked about
using them to replace traditional silicon
These breakthroughs represent a signifi-
cant step toward next-generation computer
systems far more energy efficient than any-
thing built to date.
Shulaker’s drive led him to work on another
feat: monolithic, three-dimensional nanosys-
tems. These fuse microprocessor, memory,
and additional functional layers directly on
top of one another using carbon nanotubes.
Traditional designs have the microchip and
memory on separate chips connected by
wires. But moving massive quantities of data
between those chips leads to slowdowns and
wasted energy—a problem known in the
industry as “the memory wall.” Shulaker’s
3D nanosystems solve it. —RUSS JUSKALIAN
42
AGE AFFI L I A T I O N
JINXING Institute of Plasma Physics,
ZHENG 34 Chinese Academy of Sciences
I N NOVATION
He created new physics models
for controlling fusion reactions and
hot plasma.
inxing Zheng has devised better ways to model
J the use of powerful magnets for controlling
plasma at extreme temperatures, a major
advance for fusion-based energy. Zheng’s
work is helping China leapfrog the rest of the
world and design the largest fusion reactor to date,
called the China Fusion Engineering Test Reactor.
CFETR is expected to finish construction and go
online before 2035, though it may take five to 10
years to reach full power.
Fusion reactors, based on the energy released
when atoms are combined, have great potential for
creating clean energy and are inherently safer than
existing nuclear power based on fission reactions.
But no one has built a practical one, in part because
it’s so challenging to contain the necessary plasma,
which can reach temperatures of hundreds of millions
of degrees Celsius.
Zheng’s innovation amounts to having discovered
new theoretical models for understanding how multiple
large superconducting magnets can rapidly change
their magnetic fields to keep plasma in one place
while fusion reactions occur. In 2018, with the help
of Zheng’s models, a fusion reactor in Hefei, China,
called the Experimental Advanced Superconducting
Tokamak—nicknamed “the artificial sun”—controlled
plasma at a record temperature of 50 million ˚C for
102 seconds.
China’s future CFETR is intended to operate at
over 1 gigawatt of power sometime in the 2030s. That’s
double the power of ITER, a fusion reactor currently
being completed in the south of France with cooper-
ation from countries around the world.
—RUSS JUSKALIAN
Change
AGE AFFILIATION
MARC University of
MISKIN 34 Pennsylvania
INNOVATION
He figured out how to give motion to
microscopic robots.
arc Miskin has given life to a technology
M that’s eluded the world’s top nanoscien-
tists for decades: robots too small to see.
Miskin’s tiny bots piggyback on more than
50 years of electronics innovation, making it
possible to build silicon chips smaller than the width
of a human hair. The challenge was getting these cir-
cuits, which function as the robots’ brains, to move:
previous approaches to connecting them to a pair of
microscopic legs required too much voltage to work
at such a tiny scale.
His technique fabricates legs from sheets of plat-
inum a dozen or so atoms thick, capped on one side
with an even smaller layer of titanium. When activated
with a current—generated by solar cells attached to
the robot brain—the platinum bends, causing the bot
to march forward. Miskin’s initial prototype, which
he developed as a postdoctoral researcher at Cornell
University, requires only one-fifth of a volt to move
and measures just 40 by 40 microns—smaller than
many single-celled microorganisms. It’s recognized
by Guinness World Records as the smallest ever
walking robot, and a million of them at a time can be
fabricated on a single 10-centimeter wafer.
For now, Miskin’s robot does little more than prance
COURTESY PHOTO (ZHENG); UPENN (MISKIN)
under a microscope, but his lab at the University of
Pennsylvania, where he’s a professor of electrical
and systems engineering, is fabricating limbs for a
“smart bot” with programmable memory, developed
with researchers at the University of Michigan. In the
longer term, Miskin envisions tiny bots being used
to engineer new materials, rid crops of pests, or even
act as microscopic surgeons, programmed to elimi-
nate cancer cells one by one. —JONATHAN W. ROSEN
 35 Innovators Under 35 43

Nakatsuka built her sensors using mole-

ADNAN cules called aptamers, which can be designed
MEHONIC to have strong affinity for specific targets.
AGE AFFILIATION
A GE AFFILIATIO N NAKO She first used an aptamer constructed from
34
University College London
and Intrinsic
NAKATSUKA 31 ETH Zurich DNA that changes its shape in the presence
INNOVATION
of serotonin, a neurotransmitter that plays an
I NNOVATION
important role in bodily functions like sleep
Memristors can be a Her miniature biosensors could and appetite, and in conditions like depres-
new and more efficient give scientists better insight into sion and obsessive-compulsive disorder.
building block of modern depression and dementia. Later she developed a way to attach the
computers. aptamer to the opening of a tiny pipette, just
10 nanometers in diameter, hooked up to an
electrical circuit. As the aptamer changes its
shape in the presence of serotonin, it alters
Memristors are a novel type of elec- Nako Nakatsuka is building tiny sensors that the electrical current. The sensor can measure
tric circuit element that were first can detect chemical changes in the brain and samples in brain fluid or tissue, or potentially
theorized to exist in 1971. In 2008, other parts of the body more precisely than directly next to individual neurons in a lab
researchers at Hewlett-Packard iden-
ever before. Scientists can use such informa- dish or in the brain.
tified them for the first time, in nano-
tion to help them understand and treat condi- “This might help us to better understand
devices made from titanium dioxide,
but the technology has not replaced tions like depression and dementia. Compared Parkinson’s and other diseases,” Nakatsuka
flash memory as initially predicted. with earlier sensors, Nakatsuka’s are better at says. Her sensors could be used to monitor
Resistors are elements of a cir- differentiating between structurally similar how neurons in or from a patient with such
cuit that control the flow of electric chemicals, like neurotransmitters and their a disease function in real time. And since
current. A memristor, as its name precursors and metabolites. aptamers can be used in all sorts of tests,
suggests, is like an adjustable resis-
For now, her sensors are used to take Nakatsuka’s technology could lead to faster,
tor with memory. Turn the power off,
measurements on samples in the lab, but the cheaper, and more accurate detection for all
and a memristor “remembers” the
most recent resistance it had. That technology is being refined to work directly in kinds of medical conditions and infections.
holds the promise of faster, more the body and on a wider range of chemicals. —RUSS JUSKALIAN
efficient chips that integrate mem-
ory with logic.
Adnan Mehonic is developing
memristors out of silicon oxide, the
material most commonly used in
computer chips. His most straight-
forward goal is to make dense, low-
power, high-speed memory. More
ambitiously, he is using the physics of
memristors to implement in-memory
computing and brain-like functional-
ities for future neuromorphic systems.
Among other applications, memris-
tors could greatly improve the energy
efficiency of AI systems, proponents
say. “Crossbar arrays” of memris-
tors, says Mehonic, could perform
deep-learning tasks using one-500th
as much energy as current hardware.
COURTESY PHOTO (NEHONIC); ALEXANDER TANNO (NAKATSUKA)

A startup he cofounded concluded a

$1.9 million financing round in March.
—PATRICK HOWELL O’NEILL
 44
Change

GUTTER CREDIT HERE

 35 Innovators Under 35
They’re building
companies aimed
at solving some of
the world’s most
pressing problems.
JANICE CHEN
AGE AFFILIATION
30 Mammoth Biosciences
IN N O V A T I O N
She’s using CRISPR to
make new diagnostic tests.
Janice Chen was jumping into an
Uber, cramming in equipment the
size of a microwave. At the time a
PhD student at the University of
California, Berkeley, Chen had been
invited to a lab to look for the human
papillomavirus in hospital medical
samples using a new technique she
had created.
Soon enough, bingo. Her test,
which uses the gene-editing tool
CRISPR, was able to spot the virus
nearly every time, offering a new
way to test for germs. She and sev-
eral other students, along with
Jennifer Doudna, the co-discoverer
CHRISTIE HEMM KLOK
of CRISPR, cofounded a company
with plans to develop a new gener-
ation of testing instruments. They
called it Mammoth Biosciences.
The diagnostics business isn’t
easy to break into: a few companies
with well-established technologies
dominate. Chen is now in charge of
a team of 40 as the chief technology
officer of Mammoth. She says she
leans on her experience playing chess
competitively as a teenager, when she
learned how to build a position move
by move, make meaningful sacrifices,
and get inside competitors’ minds.
Chen grew up in Salt Lake City.
Her parents were immigrants from
China. Her brother is a world cham-
pion and Olympic medalist in figure
skating. When she was growing up,
she says, her parents urged her and
her siblings to “find your passion and
do your best to move it forward in a
significant way.”
45
Chen crammed for years studying
chess moves but ended up finding
her real interest while moonlighting
at her father’s biotech supply com-
pany. That’s where she first copied
genes and engineered a bacterium.
Then, at Princeton University,
she had a chance to help out on a
large ongoing project to assemble the
entire genome of a yeast cell from
DNA parts. As an undergraduate,
she did menial lab tasks. Still, here
was life being engineered from the
ground up. And she was part of it.
For her PhD, Chen landed a spot
in Doudna’s Berkeley lab, where
CRISPR editing had been co-devel-
oped in 2012.
Chen joined a fast-paced hunt
to discover and understand even
more types of DNA editors and
harness them for new uses. She
demonstrated a way that a partic-
ular gene-editing enzyme could be
used as a diagnostic test. Her test
could find a specific sequence of
viral DNA in a sample, cut it, and
unleash a fluorescent signal that
would report the result.
That looked useful enough for
infectious-disease testing to try to
commercialize it, which is what led
her to cofound Mammoth in 2017.
Then came covid-19. When the
rollout of the standard tests stum-
bled in the spring of 2020, the US
Food and Drug Administration gave
Mammoth and dozens of other
smaller companies an emergency
green light to sell their tests for the
virus. It was a crisis, and that meant
the purse strings were loosened too.
Mammoth has won $30 million in
government funding since the pan-
demic began.
As of May 2021, Mammoth was
preparing to commercialize the com-
pany’s first product, kits that pub-
lic health labs can use to run 1,500
simultaneous covid-19 tests with less
human intervention than existing
ones require. —ANTONIO REGALADO
46
AGE AFFILIATION
JACOB Strand
BECRAFT 30 Therapeutics
I N NOVATION
He runs a company that’s
figuring out the next steps for
messenger RNA.
afe and effective covid-19 vaccines
S
have finally provided an exit to the
pandemic. The most innovative
of these vaccines use messenger
RNA—strings of nucleic acids—
to instruct cells to make a protein found
in the virus, causing the body to produce
antibodies against it. Now scientists are
eyeing all sorts of other potential uses for
this underlying technology.
“With covid-19, we have gone from
mRNA being potentially useful to say-
ing we know it works in humans,” says
Jacob Becraft. He runs a startup called
Change
Strand Therapeutics, which is work-
ing on the next step for mRNA—ways
to “program” the molecules to do addi-
tional useful tricks, like turn on only in
specific cell types, at specific times, or
automatically copy themselves so as to
strengthen their effects.
Though mRNA’s effects are tempo-
rary (because it’s an unstable molecule),
using it is in many ways simpler, safer,
and faster than trying to change the
genome of a cell. One idea the company
is pursuing is to use injections of mRNA
to instruct the body’s immune cells to
attack cancers of the skin and breast.
Becraft grew up in a small farming
community in central Illinois, famous
as home to a federal lab that discovered
how to mass-produce penicillin during
World War II. In high school, he says, he
didn’t have patience for pictures of cells,
with their labeled parts.
“It wasn’t until college that I got
exposed to biology as a machine, not
just a list of things to memorize,” he
says. “But when someone tells me how a
system works, I get it. I can imagine it.”
—ANTONIO REGALADO
TAMMY
HSU
AGE AFFILIATION
30 Huue
INNOVATION
Her new dye can
make one of the world’s
most common types of
clothing more environ-
mentally friendly.
Many consumers don’t realize
that indigo, the signature color of
denim, requires synthetic chem-
icals like formaldehyde and cya-
nide, which can be harmful to
workers and can sometimes con-
taminate local water sources.
Given that jeans are one of the
most ubiquitous clothing items in
the world, this is a huge environ-
mental problem.
Tammy Hsu, the chief scien-
tific officer of Huue, worked with
colleagues to study how color
is made in nature and program
microbes to enzymatically pro-
duce the shade they wanted. The
result is a sustainable solution that
doesn’t rely on harmful processes
or chemicals. Now the challenge
is to make the natural dye as
cheap to use as the synthetics the
industry relies upon. “The chemi-
cal industry has had 100 years to
hone their process and make it
cost efficient,” Hsu says. “We were
founded two years ago. We’re try-
ing to catch up with that. That’s
one of our biggest goals, to drive
down the price of our process.”
Huue is on track to release its
indigo dye next year. Next up for
Hsu is figuring out how to coax
microbes to produce a range of
ILLUSTRATION SOURCE: TAMMY BECRAFT; ABOLFAZL AGHANOURI
different dyes. “We’re trying to pro-
vide the fashion industry with an
alternative way,” she says.
—TANYA BASU
 35 Innovators Under 35
Judges
Pieter Abbeel
Professor, UC Berkeley;
Director, Berkeley
Robot Learning Lab
Animashree Anandkumar
Bren Professor, Caltech,
AGE AFFILIATION and Director of AI Research,
SARA Swarm Nvidia
SPANGELO 34 Technologies
David Berry
I NNOVATION CEO, Valo Health; General
Partner, Flagship Pioneering
Her tiny satellites could bring
connectivity to the remotest Ed Boyden
places on Earth. Y. Eva Tan Professor in
Neurotechnology, MIT/HHMI
Meredith Broussard
Associate Professor, NYU
Yet-Ming Chiang
Sara Spangelo didn’t quite make it as an astronaut. Kyocera Professor of Materials
Science and Engineering, MIT
But four years after an unsuccessful tryout with
Canada’s space agency, she’s achieved her own James Collins
space milestone: unveiling the world’s lowest-cost Termeer Professor, MIT
always-available satellite communications network.
John Dabiri
Spangelo, who holds a PhD in aerospace engi- Centennial Professor of
neering from the University of Michigan, is CEO Aeronautics and Mechanical
of Swarm Technologies, which seeks to provide Engineering, Caltech
affordable data services for devices anywhere on Gozde Durmus
Earth. Today, nearly 90% of the planet’s surface, Assistant Professor,
including oceans, deserts, and polar regions, lacks Stanford University
internet access. Connecting via satellite has long
Oren Etzioni
been cost-prohibitive, because satellite networks typ- CEO, Allen Institute for AI;
ically cost billions of dollars to deploy and maintain. Professor Emeritus,
Computer Science,
The key to lowering costs was to bring down
University of Washington
size: Swarm’s satellites, roughly the size of a slice
of French toast, are the smallest two-way commu- David Fattal
nication devices in orbit today. Because they’re so Founder and CEO, Leia, Inc.
compact, they can hitch rides on commercial rockets Javier Garcia Martinez
for bargain prices: total launch costs for Swarm’s full Professor of Inorganic
constellation of 150 satellites, which the company Chemistry, University of
Alicante, Spain;
will finish placing in low Earth orbit by the end of
President-elect, IUPAC
2021, will run less than $3 million.
Swarm’s data connection, which uses the VHF Julia R. Greer
radio spectrum, won’t enable seafarers to stream Mettler Professor of
Materials, Mechanics, and
Netflix: its current transfer rate of 1 kilobit per
ILLUSTRATION SOURCE: COURTESY PHOTO
Medical Engineering, Caltech
second is similar to 1990s dial-up. Swarm’s niche,
rather, is giving customers the ability to transmit Zhen Gu
Professor, Zhejiang University
small yet highly useful packets of information from
the world’s most far-flung places. This enables them Ilan Gur
to remotely monitor water supplies, detect leaks in Founder and CEO,
Activate.org
pipelines, measure soil contents, track wildlife, or
guarantee the temperature of vaccines in cold-chain Marc Lajoie
transport. —JONATHAN W. ROSEN CEO, Outpace Bio
47
Hao Li
CEO & Cofounder, Pinscreen;
Distinguished Fellow,
UC Berkeley
Zackary Lipton
Assistant Professor of
Operations Research and
Machine Learning,
Carnegie Mellon University
Zlatko Minev
Quantum physicist,
IBM Quantum; Founder,
Open Labs
Andrew Ng
Founder, DeepLearning.AI;
CEO, Landing AI;
General Partner, AI Fund
Nicole Paulk
Assistant Professor, University
of California San Francisco
Deb Raji
Fellow, Mozilla Foundation
John Rogers
Simpson/Querrey Professor
of Materials Science and Engi-
neering, Biomedical Engineer-
ing, and Neurological Surgery,
Northwestern University
Nabiha Saklayen
CEO and Cofounder, Cellino
Rachel Sheinbein
Venture Partner, Gratitude
Railroad
Mona Sloane
Sociologist, New York
University
Cyrus Wadia
Head of WW Product
Sustainability, Amazon
Jennifer West
Dean of Engineering
and Applied Sciences,
University of Virginia
Minmin Yen
CEO and Cofounder,
PhagePro
Jackie Ying
A*STAR Senior Fellow
and Director, NanoBio Lab
Alice Zhang
CEO, Verge Genomics
Ben Zhao
Neubauer Professor
of Computer Science,
University of Chicago
48 Change

THE How a small

group of
researchers
at the
margins

FIGHT
built a
movement
to wrest
the world’s
most
powerful
emerging

TO
technology
from Big
Tech’s
control.

RECLAIM
AI By Karen Hao

ILLUSTRATION BY RICARDO SANTOS

 Radical AI 49

T
Timnit Gebru never thought
a scientific paper would cause
her so much trouble.
In 2020, as the co-lead
of Google’s ethical AI team,
Gebru had reached out to
Emily Bender, a linguistics
professor at the University of
Washington, and asked to col-
laborate on research about the
troubling direction of artificial
intelligence. Gebru wanted to
identify the risks posed by large
language models, one of the
most stunning recent break-
throughs in AI research. The
models are algorithms trained
on staggering amounts of text.
Under the right conditions, they
can compose what look like
convincing passages of prose.
For a few years, tech com-
panies had been racing to build
bigger versions and integrate
them into consumer products.
Google, which invented the
technique, was already using
one to improve the relevance
of search results. OpenAI
announced the largest one,
called GPT-3, in June 2020
and licensed it exclusively to
Microsoft a few months later.
Gebru worried about how
fast the technology was being
deployed. In the paper she
wound up writing with Bender
and five others, she detailed
the possible dangers. The mod-
els were enormously costly to
create—both environmentally
(they require huge amounts
of computational power) and
financially; they were often
trained on the toxic and abu-
Clockwise from
sive language of the internet;
top left:
Raphael Gontijo Lopes, and they’d come to dominate
Deborah Raji, research in language AI, elbow-
Rediet Abebe,
Timnit Gebru, ing out promising alternatives.
Joy Buolamwini,
William Agnew.
50
Like other existing AI tech-
niques, the models don’t actu-
ally understand language. But
because they can manipulate
it to retrieve text-based infor-
mation for users or generate
natural conversation, they can
be packaged into products and
services that make tech com-
panies lots of money.
That November, Gebru
submitted the paper to a con-
ference. Soon after, Google
executives asked her to retract
it, and when she refused, they
fired her. Two months later,
they also fired her coauthor
Margaret Mitchell, the other
leader of the ethical AI team.
The dismantling of that team
sparked one of the largest con-
troversies within the AI world
in recent memory. Defenders
of Google argued that the com-
pany has the right to supervise
its own researchers. But for
many others, it solidified fears
about the degree of control that
tech giants now have over the
field. Big Tech is now the pri-
mary employer and funder of AI
researchers, including, some-
what ironically, many of those
who assess its social impacts.
Among the world’s richest
and most powerful companies,
Google, Facebook, Amazon,
Microsoft, and Apple have
made AI core parts of their
business. Advances over the
last decade, particularly in an
AI technique called deep learn-
ing, have allowed them to moni-
tor users’ behavior; recommend
news, information, and prod-
ucts to them; and most of all,
target them with ads. Last year
Google’s advertising apparatus
generated over $140 billion in
revenue. Facebook’s generated
$84 billion.
T h e c o m p a n i e s h ave
invested heavily in the tech-
nology that has brought them
such vast wealth. Google’s
parent company, Alphabet,
acquired the London-based AI
lab DeepMind for $600 million
Change
in 2014 and spends hundreds
of millions a year to support its
research. Microsoft signed a $1
billion deal with OpenAI in 2019
for commercialization rights to
its algorithms.
At the same time, tech giants
have become large investors in
university-based AI research,
heavily influencing its scientific
priorities. Over the years, more
and more ambitious scientists
have transitioned to working for
tech giants full time or adopted
a dual affiliation. From 2018
to 2019, 58% of the most cited
papers at the top two AI con-
ferences had at least one author
affiliated with a tech giant, com-
pared with only 11% a decade
earlier, according to a study by
researchers in the Radical AI
Network, a group that seeks to
challenge power dynamics in AI.
The problem is that the
corporate agenda for AI has
focused on techniques with
commercial potential, largely
ignoring research that could
help address challenges like
economic inequality and cli-
mate change. In fact, it has
made these challenges worse.
The drive to automate tasks has
cost jobs and led to the rise of
tedious labor like data clean-
ing and content moderation.
The push to create ever larger
models has caused AI’s energy
consumption to explode. Deep
learning has also created a cul-
ture in which our data is con-
stantly scraped, often without
consent, to train products like
facial recognition systems. And
recommendation algorithms
have exacerbated political
polarization, while large lan-
guage models have failed to
clean up misinformation.
It’s this situation that Gebru
and a growing movement of
like-minded scholars want
to change. Over the last five
years, they’ve sought to shift
the field’s priorities away from
simply enriching tech compa-
nies, by expanding who gets
to participate in developing
the technology. Their goal is
not only to mitigate the harms
caused by existing systems but
to create a new, more equitable
and democratic AI.
“HELLO FROM TIMNIT”
In December 2015, Gebru sat
down to pen an open letter.
Halfway through her PhD at
Stanford, she’d attended the
Neural Information Processing
Systems conference, the largest
annual AI research gathering. Of
the more than 3,700 researchers
there, Gebru counted only five
who were Black.
Once a small meeting about
a niche academic subject,
NeurIPS (as it’s now known)
was quickly becoming the big-
gest annual AI job bonanza.
The world’s wealthiest com-
panies were coming to show
off demos, throw extravagant
parties, and write hefty checks
for the rarest people in Silicon
Valley: skillful AI researchers.
That year Elon Musk
arrived to announce the non-
profit venture OpenAI. He,
Y Combinator’s then presi-
dent Sam Altman, and PayPal
cofounder Peter Thiel had put
up $1 billion to solve what they
believed to be an existential
problem: the prospect that a
superintelligence could one day
take over the world. Their solu-
tion: build an even better super-
intelligence. Of the 14 advisors
or technical team members he
anointed, 11 were white men.
While Musk was being lion-
ized, Gebru was dealing with
humiliation and harassment. At
a conference party, a group of
drunk guys in Google Research
T-shirts circled her and sub-
jected her to unwanted hugs, a
kiss on the cheek, and a photo.
Gebru typed out a scath-
ing critique of what she had
observed: the spectacle, the
cult-like worship of AI celebri-
ties, and most of all, the over-
whelming homogeneity. This
boy’s club culture, she wrote,
had already pushed talented
women out of the field. It was
also leading the entire com-
munity toward a dangerously
narrow conception of artificial
intelligence and its impact on
the world.
Google had already deployed
a computer-vision algorithm
that classified Black people as
gorillas, she noted. And the
increasing sophistication of
unmanned drones was putting
the US military on a path toward
lethal autonomous weapons. But
there was no mention of these
issues in Musk’s grand plan to
stop AI from taking over the
world in some theoretical future
scenario. “We don’t have to proj-
ect into the future to see AI’s
potential adverse effects,” Gebru
wrote. “It is already happening.”
Gebru never published her
reflection. But she realized that
something needed to change.
On January 28, 2016, she sent
an email with the subject line
“Hello from Timnit” to five
other Black AI researchers. “I’ve
always been sad by the lack of
color in AI,” she wrote. “But
now I have seen 5 of you :) and
thought that it would be cool if
we started a black in AI group
or at least know of each other.”
The email prompted a dis-
cussion. What was it about
being Black that informed their
research? For Gebru, her work
was very much a product of
her identity; for others, it was
not. But after meeting they
agreed: If AI was going to play
a bigger role in society, they
needed more Black research-
ers. Otherwise, the field would
produce weaker science—and
its adverse consequences could
get far worse.
A PROFIT-DRIVEN AGENDA
As Black in AI was just begin-
ning to coalesce, AI was hit-
ting its commercial stride. That
year, 2016, tech giants spent an
estimated $20 to $30 billion

“WE DON’T HAVE TO PROJECT INTO
THE FUTURE TO SEE AI’S POTENTIAL
ADVERSE EFFECTS.”
on developing the technology,
according to the McKinsey
Global Institute.
He a t e d by c o r p o ra t e
investment, the field warped.
Thousands more research-
ers began studying AI, but
they mostly wanted to work
on deep-learning algorithms,
such as the ones behind large
language models. “As a young
PhD student who wants to get
a job at a tech company, you
realize that tech companies are
all about deep learning,” says
Suresh Venkatasubramanian,
a computer science professor
who now serves at the White
House Office of Science and
Technology Policy. “So you shift
all your research to deep learn-
ing. Then the next PhD student
coming in looks around and says,
‘Everyone’s doing deep learning.
I should probably do it too.’”
But deep learning isn’t the
only technique in the field.
Before its boom, there was a
different AI approach known as
symbolic reasoning. Whereas
deep learning uses massive
amounts of data to teach
algorithms about meaningful
relationships in information,
symbolic reasoning focuses
on explicitly encoding knowl-
edge and logic based on human
expertise.
Some researchers now
believe those techniques
Radical AI
should be combined. The
hybrid approach would make
AI more efficient in its use of
data and energy, and give it
the knowledge and reasoning
abilities of an expert as well
as the capacity to update itself
with new information. But com-
panies have little incentive to
explore alternative approaches
when the surest way to maxi-
mize their profits is to build
ever bigger models.
In their paper, Gebru and
Bender alluded to a basic cost of
this tendency to stick with deep
learning: the more advanced AI
systems we need are not being
developed, and similar prob-
lems keep recurring. Facebook,
for example, relies heavily on
large language models for auto-
mated content moderation. But
without really understanding
the meaning behind text, those
models often fail. They reg-
ularly take down innocuous
posts while giving hate speech
and misinformation a pass.
AI-based facial recogni-
tion systems suffer from the
same issue. They’re trained on
massive amounts of data but
see only pixel patterns—they
do not have a grasp of visual
concepts like eyes, mouths,
and noses. That can trip these
systems up when they’re used
on individuals with a differ-
ent skin tone from the people
they were shown during train-
ing. Nonetheless, Amazon and
other companies have sold
these systems to law enforce-
ment. In the US, they have
caused three known cases of
police jailing the wrong per-
son—all Black men—in the
last year.
For years, many in the AI
community largely acquiesced
to Big Tech’s role in shaping
the development and impact of
these technologies. While some
expressed discomfort with the
corporate takeover, many more
welcomed the industry’s deep
well of funding.
But as the shortcomings of
today’s AI have become more
evident—both its failure to
solve social problems and the
mounting examples that it can
exacerbate them—faith in Big
Tech has weakened. Google’s
ousting of Gebru and Mitchell
further stoked the discussion
by revealing just how much
companies will prioritize profit
over self-policing.
In the immediate aftermath,
over 2,600 Google employ-
ees and 4,300 others signed a
petition denouncing Gebru’s
dismissal as “unprecedented
research censorship.” Half a
year later, research groups are
still rejecting the company’s
funding, researchers refuse to
participate in its conference
workshops, and employees are
leaving in protest.
Unlike five years ago,
when Gebru began raising
these questions, there’s now
a well-established movement
questioning what AI should be
and who it should serve. This
isn’t a coincidence. It’s very
much a product of Gebru’s own
initiative, which began with
the simple act of inviting more
Black researchers into the field.
IT TAKES A CONFERENCE
In December 2017, the new
Black in AI group hosted its first
workshop at NeurIPS. While
51
organizing the workshop, Gebru
approached Joy Buolamwini,
an MIT Media Lab researcher
who was studying commercial
facial recognition systems for
possible bias. Buolamwini had
begun testing these systems
after one failed to detect her
own face unless she donned
a white mask. She submitted
her preliminary results to the
workshop.
Deborah Raji, then an
undergraduate researcher, was
another early participant. Raji
was appalled by the culture
she’d observed at NeurIPS. The
workshop became her respite.
“To go from four or five days of
that to a full day of people that
look like me talking about suc-
ceeding in this space—it was
such important encouragement
for me,” she says.
Buolamwini, Raji, and
Gebru would go on to work
together on a pair of ground-
breaking studies about dis-
criminatory computer-vision
systems. Buolamwini and
Gebru coauthored Gender
Shades, which showed that
the facial recognition systems
sold by Microsoft, IBM, and
Chinese tech giant Megvii
had remarkably high fail-
ure rates on Black women
despite near-perfect perfor-
mance on white men. Raji
and Buolamwini then collab-
orated on a follow-up called
Actionable Auditing, which
found the same to be true for
Amazon’s Rekognition. In
2020, Amazon would agree
to a one-year moratorium on
police sales of its product, in
part because of that work.
At the very first Black in
AI workshop, though, these
successes were distant possi-
bilities. There was no agenda
other than to build commu-
nity and produce research
based on their sorely lacking
perspectives. Many onlook-
ers didn’t understand why
such a group needed to exist.
52 Change
“THESE CHANGES THAT WE’RE
FIGHTING FOR—IT’S NOT JUST FOR
MARGINALIZED GROUPS.”
Gebru remembers dismissive Venkatasubramanian and
comments from some in the Gebru also helped create the
AI community. But for others, Fairness, Accountability, and
Black in AI pointed a new way Transparency (FAccT) con-
forward. ference to create a forum for
This was true for William research on the social and
Agnew and Raphael Gontijo political implications of AI.
Lopes, both queer men con- Ideas and draft papers dis-
ducting research in com- cussed at NeurIPS affin-
puter science, who realized ity group workshops often
they could form a Queer in become the basis for papers
AI group. (Other groups that published at FAccT, which
took shape include Latinx in AI, then showcases that research
{Dis}Ability in AI, and Muslim to broader audiences.
in ML.) For Agnew, in partic- It was after Buolamwini
ular, having such a commu- presented at the first Black
nity felt like an urgent need. in AI workshop, for exam-
“It was hard to even imagine ple, that FAccT published
myself having a happy life,” Gender Shades. Along with
he says, reflecting on the lack Actionable Auditing, it then
of queer role models in the fueled several major education
field. “There’s Turing, but he and advocacy campaigns to
committed suicide. So that’s limit government use of facial
depressing. And the queer part recognition. When Amazon
of him is just ignored.” attempted to undermine the
Not all affinity group mem- legitimacy of Buolamwini’s and
bers see a connection between Raji’s research, dozens of AI
their identity and their research. researchers and civil society
Still, each group has established organizations banded together
particular expertise. Black in to defend them, foreshadowing
AI has become the intellectual what they would later do for
center for exposing algorith- Gebru. Those efforts eventu-
mic discrimination, critiquing ally contributed to Amazon’s
surveillance, and developing moratorium, which in May the
data-efficient AI techniques. company announced it would
Queer in AI has become a extend indefinitely.
center for contesting the ways The research also set off a
algorithms infringe on people’s cascade of regulation. More
privacy and classify them into than a dozen cities have
bounded categories by default. banned police use of facial
recognition, and Massachusetts
now requires police to get a
judge’s permission to use it.
Both the US and the European
Commission have proposed
additional regulation.
“First we had to just be
there,” says Gebru. “And at
some point, what Black in AI
says starts to become import-
ant. And what all of these
groups together say becomes
important. You have to listen
to us now.”
FOLLOW THE MONEY
After Gebru and Mitchell’s fir-
ing, the field is grappling anew
with an age-old question: Is it
possible to change the status
quo while working from within?
Gebru still believes working
with tech giants is the best way
to identify the problems. But
she also believes that corporate
researchers need stronger legal
protections. If they see risky
practices, they should be able
to publicly share their obser-
vations without jeopardizing
their careers.
Then there’s the question
of funding. Many researchers
want more investment from
the US government to support
work that is critical of com-
mercial AI development and
advances the public welfare.
Last year, it committed a mea-
sly $1 billion to non-defense-
related AI research. The Biden
administration is now asking
Congress to invest an addi-
tional $180 billion in emerging
technologies, with AI as a top
priority.
Such funding could help
people like Rediet Abebe, an
assistant professor of computer
science at the University of
California, Berkeley. Abebe
came into AI with ideas of using
it to advance social equity. But
when she started her PhD at
Cornell, no one was focused
on doing such research.
In the fall of 2016, as a PhD
student, she began a small
Cornell reading group with a
fellow graduate student to study
topics like housing instability,
health-care access, and inequal-
ity. She then embarked on a
new project to see whether her
computational skills could sup-
port efforts to alleviate poverty.
Eventually, she found
the Poverty Tracker study, a
detailed data set on the finan-
cial shocks—unexpected
expenses like medical bills or
parking tickets—experienced
by more than 2,000 New York
families. Over many conversa-
tions with the study’s authors,
social workers, and nonprofits
serving marginalized commu-
nities, she learned about their
needs and told them how she
could help. Abebe then devel-
oped a model that showed
how the frequency and type
of shocks affected a family’s
economic status.
Five years later, the project
is still ongoing. She’s now col-
laborating with nonprofits to
improve her model and work-
ing with policymakers through
the California Policy Lab to
use it as a tool for preventing
homelessness. Her reading
group has also since grown
into a 2,000-person commu-
nity and is holding its inaugu-
ral conference later this year.
Abebe sees it as a way to
incentivize more researchers
to flip the norms of AI. While
traditional computer science
conferences emphasize advanc-
ing computational techniques
for the sake of doing so, the
new one will publish work that
first seeks to deeply understand
a social issue. The work is no
less technical, but it builds the
foundation for more socially
meaningful AI to emerge.
“These changes that we’re
fighting for—it’s not just for
marginalized groups,” she says.
“It’s actually for everyone.” Q
Karen Hao is a senior
editor for AI at MIT
Technology Review.
CHANGE
PRESENTS
OPPORTUNITIES
Make decisions that make a
difference with the Financial Times
Read more at ft.com/newagenda
 THE
54

A year into the covid-19 pandemic, Apple commem-

GREAT
orated the growing array of devices featuring its cus-
tom M1 chip with great fanfare, including a “Mission
Implausible” ad on TV featuring a young man run-
ning across the rooftops of its “spaceship” campus
in Cupertino and infiltrating the facility to “steal” the
breakthrough microprocessor from a MacBook and
place it inside an iPad Pro.
Apple’s custom-designed chip is the latest triumph
for Moore’s Law, the observation turned self-fulfilling
prophecy that chipmakers can double the number of
transistors on a chip every few years. The M1 packs

CHIP
16 billion transistors on a microprocessor the size of
a large postage stamp. It’s a marvel of today’s semi-
conductor manufacturing prowess.
But even as Apple celebrated the M1, the world
was facing an economically devastating shortage of
microchips, particularly the relatively cheap ones that
make many of today’s technologies possible.
Automakers have been shutting down assembly
lines and laying off workers because they can’t get
enough $1 chips. Manufacturers have resorted to
building vehicles without the chips necessary for
navigation systems, digital rear-view mirrors, display

DDIV
touch screens, and fuel management systems. Overall,
the global automotive industry could lose more than
$110 billion to the shortage in 2021.
Production has also slowed for smartphones,
laptops, video-game consoles, TVs, and even smart
appliances, all because of the lack of cheap micro-
chips. Their use is so essential and so widespread that
some observers think the chip crisis could threaten
the global economic recovery from the pandemic.
The global shortage is shining a harsh spotlight on
the semiconductor industry’s ability to deliver cheaper
A shortage of microchips
and more powerful microchips. The longstanding
is threatening to slow digital
promise of chips with ever more capabilities inspired
innovation inspired by
engineers, programmers, and product designers to
the promise of ever faster,
create generations of new products and services.
cheaper computing power.
Moore’s Law has been more than just a road map for
the semiconductor industry—it has governed techno-
logical change over the last half-century. By Jeremy Hsu
Now that promise of more computing power every-
where is crumpling, but not because chipmakers have
COURTESY OF ASML
55

This tool, called an extreme

ultraviolet lithography machine,
allows manufacturers to make
leading-edge chips with features a
few nanometers in size.
56 Change

finally run up against the physical limits of technology to make now beyond the reach of many companies. In addition to Apple,
ever smaller transistors. Instead, the growing costs of sustaining only the largest tech companies that require the highest com-
Moore’s Law have encouraged consolidation among chipmak- puting performance, such as Qualcomm, AMD, and Nvidia, are
ers and created more choke points in the immensely complex willing to pay hundreds of millions of dollars to design a chip
business of chip production. for leading-edge nodes, says Sri Samavedam, senior vice presi-
Even as microchips have become essential in so many prod- dent of CMOS technologies at Imec, an international research
ucts, their development and manufacturing have come to be institute based in Leuven, Belgium.
dominated by a small number of producers with limited capac- Many more companies are producing laptops, TVs, and cars
ity—and appetite—for churning out the commodity chips that that use chips made with older technologies, and a spike in
are a staple for today’s technologies. And because making chips demand for these is at the heart of the current chip shortage.
requires hundreds of manufacturing steps and months of pro- Simply put, a majority of chip customers can’t afford—or don’t
duction time, the semiconductor industry cannot quickly pivot want to pay for—the latest chips; a typical car today uses doz-
to satisfy the pandemic-fueled surge in demand. ens of microchips, while an electric vehicle uses many more. It
After decades of fretting about how we will carve out fea- quickly adds up. Instead, makers of things like cars have stuck
tures as small as a few nanometers on silicon wafers, the spirit with chips made using older technologies
of Moore’s Law—the expectation that cheap, powerful chips will What’s more, many of today’s most popular electronics sim-
be readily available—is now being threatened by something far ply don’t require leading-edge chips. “It doesn’t make sense to
more mundane: inflexible supply chains. put, for example, an A14 [iPhone and iPad] chip in every single
computer that we have in the world,” says
A lonely frontier Hassan Khan, a former doctoral researcher
Twenty years ago, the world had 25 manu- at Carnegie Mellon University who stud-
facturers making leading-edge chips. Today, ONLY THE LARGEST ied the public policy implications of the
only Taiwan Semiconductor Manufacturing TECH end of Moore’s Law and currently works
COMPANIES ARE
Company (TSMC) in Taiwan, Intel in the WILLING
at Apple. “You don’t need it in your smart
United States, and Samsung in South Korea TO PAY HUNDREDS OF thermometer at home, and you don’t need
have the facilities, or fabs, that produce MILLIONS OF DOLLARS 15 of them in your car, because it’s very
TO DESIGN
the most advanced chips. And Intel, long A CHIP FOR power hungry and it’s very expensive.”
a technology leader, is struggling to keep LEADING-EDGE The problem is that even as more users
NODES.
up, having repeatedly missed deadlines for rely on older and cheaper chip technol-
producing its latest generations. ogies, the giants of the semiconductor
One reason for the consolidation is industry have focused on building new
that building a facility to make the most leading-edge fabs. TSMC, Samsung, and
advanced chips costs between $5 billion and $20 billion. These Intel have all recently announced billions of dollars in invest-
fabs make chips with features as small as a few nanometers; in ments for the latest manufacturing facilities. Yes, they’re expen-
industry jargon they’re called 5-nanometer and 7-nanometer sive, but that’s where the profits are—and for the last 50 years,
nodes. Much of the cost of new fabs goes toward buying the it has been where the future is.
latest equipment, such as a tool called an extreme ultraviolet TSMC, the world’s largest contract manufacturer for chips,
lithography (EUV) machine that costs more than $100 million. earned almost 60% of its 2020 revenue from making leading-edge
Made solely by ASML in the Netherlands, EUV machines are used chips with features 16 nanometers and smaller, including Apple’s
to etch detailed circuit patterns with nanometer-size features. M1 chip made with the 5-nanometer manufacturing process.
Chipmakers have been working on EUV technology for more Making the problem worse is that “nobody is building semi-
than two decades. After billions of dollars of investment, EUV conductor manufacturing equipment to support older technolo-
machines were first used in commercial chip production in 2018. gies,” says Dale Ford, chief analyst at the Electronic Components
“That tool is 20 years late, 10x over budget, because it’s amazing,” Industry Association, a trade association based in Alpharetta,
says David Kanter, executive director of an open engineering Georgia. “And so we’re kind of stuck between a rock and a hard
consortium focused on machine learning. “It’s almost magical spot here.”
that it even works. It’s totally like science fiction.”
Such gargantuan effort made it possible to create the bil- Low-end chips
lions of tiny transistors in Apple’s M1 chip, which was made by All this matters to users of technology not only because of the
TSMC; it’s among the first generation of leading-edge chips to supply disruption it’s causing today, but also because it threat-
rely fully on EUV. ens the development of many potential innovations. In addition
Paying for the best chips makes sense for Apple because these to being harder to come by, cheaper commodity chips are also
chips go into the latest MacBook and iPhone models, which sell becoming relatively more expensive, since each chip genera-
by the millions at luxury-brand prices. “The only company that is tion has required more costly equipment and facilities than the
actually using EUV in high volume is Apple, and they sell $1,000 generations before.
smartphones for which they have insane margin,” Kanter says. Some consumer products will simply demand more powerful
Not only are the fabs for manufacturing such chips expen- chips. The buildout of faster 5G mobile networks and the rise
sive, but the cost of designing the immensely complex circuits is of computing applications reliant on 5G speeds could compel
Read
smarter,
not
harder
A subscription to MIT Technology Review includes

• Unlimited web access • Subscriber-only app

• Exclusive, subscriber- • The Algorithm newsletter
only stories • Access to 120+ years
• Digital version of each issue of publication archives

Subscribe today at technologyreview.com/subscribe

58 Change

investment in specialized chips designed for networking equip- 13%. Taiwan’s TSMC alone has nearly 55% of the foundry market
ment that talks to dozens or hundreds of Internet-connected that handles consumer chip manufacturing orders.
devices. Automotive features such as advanced driver-assistance Looming over everything is the US-China rivalry. China’s
systems and in-vehicle “infotainment” systems may also benefit national champion firm SMIC has been building fabs that are
from leading-edge chips, as evidenced by electric-vehicle maker still five or six years behind the cutting edge in chip technolo-
Tesla’s reported partnerships with both TSMC and Samsung on gies. But it’s possible that Chinese foundries could help meet the
chip development for future self-driving cars. global demand for chips built on older nodes in the coming years.
But buying the latest leading-edge chips or investing in spe- “Given the state subsidies they receive, it’s possible Chinese
cialized chip designs may not be practical for many companies foundries will be the lowest-cost manufacturers as they stand
when developing products for an “intelligence everywhere” future. up fabs at the 22-nanometer and 14-nanometer nodes,” Khan
Makers of consumer devices such as a Wi-Fi-enabled sous vide says. “Chinese fabs may not be competitive at the frontier, but
machine are unlikely to spend the money to develop specialized they could supply a growing portion of demand.”
chips on their own for the sake of adding even fancier features, The global semiconductor industry will need to almost double
Kanter says. Instead, they will likely fall back on whatever chips overall capacity by 2030 to keep pace with demand, according
made using older technologies can provide. to the Semiconductor Industry Association (SIA), a Washington-
And lower-cost items such as clothing, he says, have “razor- based industry group, which has advocated for strengthening
thin margins” that leave little wiggle room for more expensive the global supply chain rather than attempting to build fully
chips that would add a dollar—let alone $10 or $20—to each “self-sufficient” domestic manufacturing capability.
item’s price tag. That means the climbing But in a nod to the importance of
price of computing power may prevent advanced chips for national security and
the development of clothing that could, critical infrastructure, the SIA suggests
for example, detect and respond to voice THE MAJORITY that the US provide “market-driven incen-
commands or changes in the weather. OF TODAY’S CHIP tives” for companies to build two or three
CUSTOMERS
The world can probably live without MAKE DO WITH THE
new leading-edge fabs domestically. That
fancier sous vide machines, but the lack CHEAPER COMMODITY could help ensure that the nation’s core
of ever cheaper and more powerful chips CHIPS THAT telecommunications networks and data
REPRESENT
would come with a real cost: the end of an A TRADE-OFF centers—along with the US military—have
era of inventions fueled by Moore’s Law BETWEEN COST AND a domestic supply of chips.
PERFORMANCE.
and its decades-old promise that increas- The White House photo ops with the
ingly affordable computation power will president called to mind the role that the
be available for the next innovation. government has played since the dawn
The majority of today’s chip custom- of the semiconductor industry that gave
ers make do with the cheaper commodity chips that represent Silicon Valley its name. “Making that front and center is not some-
a trade-off between cost and performance. And it’s the supply thing that the president has talked about in that way since Ronald
of such commodity chips that appears far from adequate as the Reagan,” says Margaret O’Mara, a historian at the University
global demand for computing power grows. of Washington in Seattle. “Biden sitting there waving a wafer
“It is still the case that semiconductor usage in vehicles is around—I don’t think I’ve seen that in a presidential hand ever.”
going up, semiconductor usage in your toaster oven and for The US government became “the Valley’s first, and perhaps
all kinds of things is going up,” says Willy Shih, a professor of its greatest, venture capitalist,” O’Mara wrote in her 2019 book
management practice at Harvard Business School. “So then the The Code: Silicon Valley and the Remaking of America. Large
question is, where is the shortage going to hit next?” government orders for chips to supply NASA’s Apollo program
and the military’s Minuteman intercontinental ballistic missiles
A global concern encouraged chipmakers to begin mass production and helped
In early 2021, President Joe Biden signed an executive order lower the cost of the first silicon chips from $1,000 each in 1960
mandating supply chain reviews for chips and threw his support to just $25 by 1965.
behind a bipartisan push in Congress to approve at least $50 bil- The price drop made computing power affordable to many
lion for semiconductor manufacturing and research. Biden also beyond just deep-pocketed government agencies. It kick-started
held two White House summits with leaders from the semicon- the golden age of Moore’s Law, in which customers reaped the
ductor and auto industries, including an April 12 meeting during benefits of cheaper chips that also delivered better performance
which he prominently displayed a silicon wafer. every few years. And you might not know its promise was in peril
The actions won’t solve the imbalance between chip demand if all you had to go on was Apple’s latest ad.
and supply anytime soon. But at the very least, experts say, today’s While I was interviewing O’Mara for this story, a delivery
crisis represents an opportunity for the US government to try person showed up at her door as if on cue.
to finally fix the supply chain and reverse the overall slowdown “Speaking of chips, I’m just getting handed my brand-new
in semiconductor innovation—and perhaps shore up the US’s computer,” she said with a laugh. “Yes, I’ve got my new MacBook
capacity to make the badly needed chips. with my M1 chip.”
An estimated 75% of all chip manufacturing capacity was based Jeremy Hsu is a technology and science journalist based
in East Asia as of 2019, with the US share sitting at approximately in New York City.
Don’t
believe
everything
you hear.
The award-winning podcast
In Machines We Trust thoughtfully
examines the far-reaching impact of
artificial intelligence on our daily lives.

Download it wherever you listen.

 61

One day in late March, People’s

THOSE
Daily, the Chinese Communist
Party’s official newspaper, shared
a pair of photos on Chinese social media.
The first, in black and white, was of the signing of
the Boxer Protocol, a 1901 treaty between the Qing
empire, which ruled China at the time, and 11 foreign
nations. Troops from eight of these countries, including

WHO
the US, had occupied Beijing following sieges on their
embassies by a peasant militia known as the Boxers.
Among a litany of concessions, the Qing government
agreed to pay the eight occupying powers an indem-
nity of 450 million taels of silver (about $10 billion in
today’s dollars), almost twice its annual revenue. The

FALL
Boxer Protocol is etched into the Chinese conscious-
ness as a searing reminder of the country at its weakest.
The second image, in vivid color, was from the pre-
vious day, at an acrimonious summit held in Alaska
between top Chinese and American officials. It was
the first high-level meeting between the two govern-
ments during the Biden administration. The officials

BEHIND
criticized one another’s governments for human rights
abuses and belligerence on the international stage. At
the end of the opening session, Yang Jiechi, director
RICHARDS PHOTOGRAPHER; FROM THE SHELBY WHITE AND LEON LEVY ARCHIVES CENTER, INSTITUTE FOR ADVANCED STUDY, PRINCETON, NJ, USA (BLACKBOARD);

of foreign affairs for the Chinese Communist Party,

OPPOSITE: GETTY IMAGES (BRANCH,COUPLES); PEABODY ESSEX MUSEUM VIA WIKIMEDIA; PUBLIC DOMAIN VIA WIKIMEDIA (ENGRAVING, TICKET, STAMP); ALAN

scolded his American counterparts: “Haven’t we, the

Chinese people, suffered from foreign bullies long

GET
enough? Haven’t we been penned in by foreign nations
and stopped from progress long enough?”
The People’s Daily post quoted Yang as saying
further: “You, the United States, are not qualified to
claim that you are speaking to China from a position of
strength.” This struck a nerve; the post has been liked
almost 2 million times, and Yang’s quote has found its

BEATEN
way to T-shirts, stickers, and cell-phone covers sold in
China. To many in the country, the harsh words carry
the sweet taste of revenge. China is finally strong enough
to stand up to the most powerful nation on earth and
demand to be treated as its equal.
From the last Chinese empire to the current People’s

UP
Republic, generations of politicians and intellectu-
als have sought ways to build a strong China. Some
imported tools and ideas from the West. Others left
China for a better education, but the homeland still
beckoned. They pondered the relationships between
East and West, tradition and modernity, national alle-
giance and cosmopolitan ideals. Their accomplishments
and regrets have shaped the path of China’s develop- Can science
ment and mapped the contours of Chinese identity. build a strong
I’m a product of their complex legacy. I grew up in China?
Hefei, a medium-sized city in central-eastern China.
The Hefei of my childhood was a humble place, known
for ancient battlegrounds, sesame snacks, and a few By
good universities. I spent the first 19 years of my life Yangyang Cheng
62
there and left in 2009 to pursue my PhD in physics in
the US, where I now live and work. Watching my birth
country’s ascent conjures up mixed feelings. I’m glad
that the majority of Chinese people enjoy a higher stan-
dard of living. I’m also alarmed by the hardened edge
to China’s new superpower status. Economic growth
and technological advancements have not ushered in
more political freedoms or a more tolerant society. The
Chinese government has become more authoritarian
and its people more nationalistic. The world feels more
fractured today.
Hefei is now a budding metropolis with new research
centers, manufacturing plants, and technology startups.
For two of the city’s proudest sons, born a century apart,
a strong homeland armed with science and technology
was the aspiration of a lifetime. One of these men was
the late Qing’s most revered statesman. The other is
one of the first two Nobel laureates from China. The
Boxer Protocol marked the end of one career and laid
the foundation for the other. I grew up with their names
and have been returning to their stories. They teach me
about the forces that propelled China’s rise, the way
lives can be squeezed by the pressures of geopolitics,
and the risks of using science for state power.
IN 1823,
Li Hongzhang was born to a wealthy household in Hefei,
then a small provincial capital surrounded by farmland.
Like his father and brother before him, Li excelled in
the imperial exams, China’s centuries-old system for
selecting officials. Over six feet tall and with a pierc-
ing gaze, he commanded space and attention. He dis-
tinguished himself in suppressing peasant rebellions
and rose quickly in the imperial court to become the
Qing empire’s highest-ranking governor, its commerce
minister, and its de facto foreign minister.
After China lost to British and French forces in the
Opium Wars, Li and his allies launched a wide range
of reforms. They called it the Movement for Western
Affairs, also known as “self-strengthening.” The strategy
was best summed by the scholar Wei Yuan in an 1844
book, Illustrated Treatise on the Maritime Kingdoms:
“Learn advanced technologies from the barbarians to
keep barbarian invaders at bay.”
To the Chinese literati, the world was divided
between hua, the homeland of civilized glory, and yi,
the places where barbarians dwelled. British gunboats
on the southern shore had shaken but not shattered
this centuries-old belief. Proponents of self-strength-
ening claimed that Chinese tradition was the base
onto which Western technology could be grafted for
practical use. As the historian Philip Kuhn has argued,
such logic also implied that technology was culturally
neutral and could be detached from political systems.
A classically trained scholar and battle-tested general,
Li championed both civilian and military enterprises. He
Change
petitioned the emperor to construct the first Chinese
railroad and founded the country’s first privately owned
steamship company. He allocated generous government
funding for the Beiyang Fleet, China’s first modern navy.
In 1865, Li oversaw the establishment of the Jiangnan
Arsenal, the largest weapons factory in East Asia at the
time. In addition to producing advanced machinery for
war, the arsenal also included a school and a translation
bureau, which translated scores of Western textbooks
on science, engineering, and mathematics, establish-
ing the vocabulary in which these subjects would be
discussed in China.
Li also supervised China’s first overseas education
program, which sent a cohort of Chinese boys aged 10
to 16 to San Francisco in the summer of 1872. After a
promising start, the mission was derailed by anti-Chi-
nese racism in the US and conservative obstruction at
home. Some students, upon returning to China, were
held and questioned by the authorities about their
loyalty. After nine bumpy years, the program was shut
down in 1881 on the eve of the Chinese Exclusion Act.
Meanwhile, neighboring Japan had adopted not only
the West’s technology but also its governing methods,
transforming a feudal society into a modern industrial
state with a formidable military. For centuries, the
Chinese elite had looked down upon Japan, dismissing
it as small and inferior. When the two countries went
to war in 1894, ostensibly over the status of Korea, the
real prize was status as the preeminent Asian power.
Japan won decisively. It was six years after this devas-
tating loss that Li signed the Boxer Protocol on behalf
of the Qing government. He died two months later.
By the start of the 20th century, the last Chinese
empire had lost its legitimacy. Armed rebellions were
erupting across the country. The Qing regime was
overthrown in 1911, and the Republic of China was
born. Progressive intellectuals saw Chinese tradition
as “rotten and decayed,” a cultural albatross holding
their country back. They believed that national sal-
vation demanded embracing Western ideas. The few
dissenting voices were sidelined.
China’s path to westernization received some early
assistance from the US. Hoping to improve relations
between the two countries, the US government decided
to return almost half the American portion of the indem-
nity China had agreed to pay in the Boxer Protocol. With
the US side dictating the terms, part of the remittance
went toward a program known as the Boxer Indemnity
Scholarships, which provided one of the few pathways
for Chinese students to study in the US. The bulk of
the returned payment was used to establish a Western-
style preparatory school, which became Tsinghua
University, China’s premier technological institution.
Li Hongzhang could not have imagined that after his
death, the most shameful chapter of his career would,
at the whimsical hand of geopolitics, contribute to his
lifelong dream of bringing Western science and educa-
tion to China. Tsinghua took its motto from the ancient
 63

after the Communist victory, and pioneered the field

text of I Ching, the Book of Changes: “The
work of self-strengthening is ceaseless. The
virtuous carry the world with generosity.”
IN 1945,
a young man named Chen Ning Yang grad-
uated from Tsinghua and arrived at the University of
Chicago for his PhD on a Boxer Indemnity Scholarship.
Inspired by the autobiography of Benjamin Franklin,
which he had read as a child, the aspiring physicist from
Hefei gave himself the English name Frank.
After World War II ended, Nationalists and
Communists continued to battle in China. Yang and
his small cohort of overseas Chinese students faced
a pressing dilemma: Should they stay in the West—
despite its racism and anticommunist paranoia—and
enjoy social stability, material comfort, and career
opportunities? Or should they return to their impov-
erished homeland after graduation and help it rebuild?
In a long letter to Yang in 1947, his college classmate
Huang Kun wrote, “It’s difficult to imagine how intellec-
tuals like us can affect the fate of a nation. Independent
PEOPLE’S DAILY VIA WEIBO
of semiconductor physics in the country. Deng Jiaxian,
Yang’s best friend since adolescence, boarded a ship
back nine days after receiving his PhD from Purdue. He
became a leader in China’s fledgling nuclear weapons
program. Some overseas Chinese scientists, dreading
Communist rule, followed the Nationalist government
to Taiwan, Yang’s former mentor Wu Ta-You among
them. But Yang opted to stay in the US after getting his
doctorate, moving in 1949 to the Institute for Advanced
Study in Princeton, New Jersey. There he would spend
the better part of the next two decades. He would not
see any of his old friends for many years.
In 1957, Yang and Tsung-Dao Lee, a fellow Chinese
graduate of the University of Chicago, won the Nobel
Prize for proposing that when some elementary par-
ticles decay, they do so in a way that distinguishes
left from right. They were the first Chinese laureates.
Speaking at the Nobel banquet, Yang noted that the
prize had first been awarded in 1901, the same year as
the Boxer Protocol. “As I stand here today and tell you
about these, I am heavy with an awareness of the fact
that I am in more than one sense a product of both
the Chinese and Western cultures, in harmony and in
conflict,” he said.
Yang became a US citizen in 1964 and moved to
Stony Brook University on Long Island in 1966 as
the founding director of its Institute for Theoretical
Physics, which was later named after him. As
People’s Daily the relationship between the US and China
juxtaposed a
began to thaw, Yang visited his homeland in
photograph from the
signing of the Boxer 1971—his first trip in a quarter of a century. A
Protocol with one from lot had changed. His father’s health was fail-
the Alaska summit.
The post has been ing. The Cultural Revolution was raging, and
liked almost 2 million both Western science and Chinese tradition
times.
had been deemed heresy. Many of Yang’s for-
mer colleagues, including Huang and Deng,
were persecuted and forced to perform hard labor. The
Nobel laureate, on the other hand, was received like a
foreign dignitary. He met with officials at the highest
levels of the Chinese government and advocated for
the importance of basic research.
In the years that followed, Yang visited China regu-
larly. At first, his trips drew attention from the FBI, which
saw exchanges with Chinese scientists as suspect. But
by the late 1970s, hostilities had waned. Mao Zedong
was dead. The Cultural Revolution was over. Beijing
adopted reforms and opening-up policies. Chinese
students could go abroad for study. Yang helped raise
funding for Chinese scholars to come to the US and for
international experts to travel to conferences in China,
where he also helped establish new research centers.
When Deng Jiaxian died in 1986, Yang wrote an emo-

minds like us, once we go back, will certainly get crushed
like grains in a mill … but I still sincerely believe that
whether China has us makes a difference.”
Huang was studying in England at the University
of Bristol. He returned to China in 1951, two years
tional eulogy for his friend, who had devoted his life to
China’s nuclear defense. It concluded with a song from
1906, one of his father’s favorites: “[T]he sons of China,
they hold the sky aloft with a single hand … The crim-
son never fades from their blood spilled in the sand.”
64 Change

Yang retired from Stony Brook in 1999 and moved on academic exchange between the two countries. The
back to China a few years later to teach freshman physics US government has also placed various restrictions on
at Tsinghua. In 2015, he renounced his US citizenship Chinese students, shortening their visas and denying
and became a citizen of the People’s Republic of China. access to facilities in disciplines deemed “sensitive.”
In an essay remembering his father, Yang recounted There are real problems of illicit behavior in Chinese
his earlier decision to emigrate. He wrote, “I know that talent programs. Earlier this year, a chemist associated
until his final days, in a corner of his heart, my father with Thousand Talents was convicted in Tennessee of
never forgave me for abandoning my homeland.” stealing trade secrets for BPA-free beverage can liners.
A hospital researcher in Ohio pled guilty to stealing
designs for exosome isolation used in medical diagnosis.
IN 2007,
Some US-based scientists failed to disclose additional
income from China in federal grant proposals or on tax
when he was 85 years old, Yang stopped by our home- returns. All these are cases of individual greed or neg-
town on an autumn day and gave a talk at my university. ligence. Yet the FBI considers them part of a “China
My roommates and I waited outside the venue hours threat” that demands a “whole-of-society” response.
in advance, earning precious seats in the packed audi- The Biden administration is reportedly considering
torium. He took the stage to thunderous applause and changes to the China Initiative, which many science
delivered a presentation in English about his Nobel- associations and civil rights groups have criticized as
winning work. I was a little perplexed by his choice of “racial profiling.” But no official announcements have
language. One of my roommates muttered, wondering been made. New cases have opened under Biden;
whether Yang was too good to speak in his mother restrictions on Chinese students remain in effect.
tongue. We listened attentively nevertheless, grateful Seen from China, the sanctions, prosecutions, and
to be in the same room as the great scientist. export controls imposed by the US look like continu-
A college junior and physics major, I was preparing to ations of foreign “bullying.” What has changed in the
apply to graduate school in the US. I’d been raised with past 120 years is China’s status. It is now not a crum-
the notion that the best of China would leave China. bling empire but a rising superpower. Policymakers in
Two years after hearing Yang in person, I too enrolled
at the University of Chicago. I received my PhD in
2015 and stayed in the US for postdoctoral research.
Months before I bid farewell to my homeland, the
central government launched its flagship overseas
recruitment program, the Thousand Talents Plan,
encouraging scientists and tech entrepreneurs to move
to China with the promise of generous personal com-
pensation and robust research funding. In the decade
since, scores of similar programs have sprung up. Some,
like Thousand Talents, are supported by the central gov-
ernment. Others are financed by local municipalities.
Beijing’s aggressive pursuit of foreign-trained talent
is an indicator of the country’s new wealth and tech-
nological ambition. Though most of these programs
are not exclusive to people of Chinese origin, the pro-
motional materials routinely appeal to sentiments of
national belonging, calling on the Chinese diaspora to
come home. Bold red Chinese characters headlined the
web page for the Thousand Talents Plan: “The moth-
erland needs you. The motherland welcomes you. The
motherland places her hope in you.”
These days, though, the website isn’t accessible.
W. & D. DOWNEY, PUBLIC DOMAIN, VIA WIKIMEDIA

Since 2020, mentions of the Thousand Talents Plan have

largely disappeared from the Chinese internet. Though
the program continues, its name is censored on search
engines and forbidden in official documents in China.
Since the final years of the Obama administration, the
Chinese government’s overseas recruitment has come
under intensifying scrutiny from US law enforcement. In
2018, the Justice Department started a China Initiative
intended to combat economic espionage, with a focus
 65

It was a very difficult decision, and I’m still reckoning

with the sense of loss associated with the change. But
with every passing day, news from my birth country
and my adopted home reminds me of why I made the
choice. Advancements in science and technology have
created unprecedented wealth—as well as inequality and
capacity to cause harm. In the fevered race for power
and supremacy, concerns about ethics and sustainabil-
ity are drowned out by jingoistic cheers.
My mother has been trying to persuade me to
move back to China. She tells me how Hefei has shed
its rusty, blue-collar image to become a modern city.
“It has a new subway system! Do you know how fast it
is?” she says over the phone. The sincerity in her voice
breaks my heart.
I want to say that I do not care for fast trains or new
buildings—I really don’t—but I also know that my
mother does not care for these things either. Her pride
in her country’s development is genuine. If there’s any-
thing she loves more than her homeland, though, it’s her
child. My mother wants me to come back not because
of some lofty ideals of patriotism, though she believes
in them; nor for my career advancement, though the
Chinese government has been investing heavily in the
fundamental sciences. My mother wants me to come
back because she is afraid.
My mother is afraid that the borders between the
US and China will be closed again as they were during
the pandemic, shut down by forces just as
countries use similar techno-nationalistic lan- Above: invisible as a virus and even more deadly. She
Yang (seated, left)
guage to describe science as a tool of national fears for my safety in a foreign land that is in
with fellow Nobel
greatness and scientists as strategic assets in Prize winners (clock- many ways increasingly hostile to my race and
geopolitics. Both governments are pursuing wise from left) Val nationality. What my mother does not know,
Fitch, James Cronin,
military use of technologies like quantum Samuel C.C. Ting, and or refuses to accept, is that the homeland is
computing and artificial intelligence. Isidor Isaac Rabi not safe for me either. A state can command
“We do not seek conflict, but we welcome Opposite:
the world’s second-largest economy and a
stiff competition,” National Security Advisor Li (center) pictured strong military, and still be too fragile to allow
Jake Sullivan said at the Alaska summit. Yang with Lord Salisbury dissent. Sometimes, life as a Chinese person
(left) and Lord Curzon
Jiechi responded by arguing that past confron- (right) during a trip means following one’s conscience with no
tations between the two countries had only to England in 1896 refuge in sight.
damaged the US, while China pulled through. At Li Hongzhang’s family temple on the
Much of the Chinese public relishes the prospect outskirts of Hefei, there is an old yulan tree. Tall and
of competing against the US. Take a popular saying fragrant, yulan was a favorite of royalty. Legend has
of Mao’s: “Those who fall behind will get beaten up!” it that this tree was a gift from the Japanese prime
The expression originated from a speech by Joseph minister on Li’s 70th birthday. Li planted it himself.
Stalin, who stressed the importance of industrialization In less than a year, the two countries would be at war.
for the Soviet Union. For the Chinese public, largely The tree has outlived both men and the empires they
unaware of its origins, it evokes the recent past, when served. It blossoms every year and occasionally bears
a weak China was plundered by foreigners. When I fruit. It is a witness, and also a teacher. One day, when
ENERGY.GOV, PUBLIC DOMAIN, VIA WIKIMEDIA

was little, my mother often repeated the expression at
home, distilling a century of national humiliation into
a personal motivation for excellence. It was only later,
in adulthood, that I began to question the underlying
logic: Is a competition between nations meaningful?
By what metric, and to what end?
After 11 years of designing particle detectors and
searching for dark matter, I left physics at the end of
2020 for a position working on science policy and ethics.
I’m able to go back to China and to Hefei, I hope to
visit Li’s old residence.
I hope to be there in the spring, when the yulan
blooms. Its flowers will be the purest white. Its petals
will be thick and smooth. Its branches will lift into the
sky. When the sun hits at just the right spot, its shadow
will carry the shape of home.
Yangyang Cheng is a particle physicist and a
postdoctoral fellow at Yale Law School.
 67

Opposite: A mud core from Fairy Lake in Montana, superimposed against the surrounding mountains, reveals
thousands of years of vegetative history. Geographer James Benes annotated the photo.
FAST TIME
Climate change is warping geological time, compressing the time scales of
natural processes. In photographs taken around the world, Ian van Coller
has documented these shifts, reflected in rocks, sediment, and the shrinking
of glaciers. Van Coller collaborates with scientists who annotate his images,
pointing out key geological features. He also uses historical photos to show
changes, juxtaposing the black-and-white images taken by earlier expedi-
tions with today’s landscapes; peaks once covered in snow are now bare rock.
 Quelccaya Glacier in Peru, seen here in 2017, is receding. The foreground rocks show signs of glacial erosion
and were likely still covered 10 years ago. Each layer in the ice represents a year’s worth of snow. Annotated by geographer Carsten Braun.
68
70
 A photograph taken in 2020 shows just how little is left of the glacier at Mount Stanley in Uganda.
The photo from a 1906 expedition shows the glacier below Elena Peak; today what’s left is dirty ice, a sign the glacier will soon be gone. (Carsten Braun)
71
72
 At nearby Mount Baker, also in Uganda, the story is similar. Dotted lines are an attempt to estimate the ice seen in
earlier expeditions on Semper Peak, which is now bare rock. There is no sign of what the 1906 photo labels Moore Glacier. (Carsten Braun)
73
 RE
“I have a stalker—an online harasser who has
moved into real-world stalking—which made me
more fearful. ” —p. 87

VIEW
Books,
policy, and culture
in perspective

SARAH JAFFE
he workers at the Amazon fulfillment center
T in Bessemer, Alabama, wanted a union.
Power play The center opened in March of last year,
just as stay-at-home orders for covid-19 went into
Workers across the technology industry are organizing in effect. While much of the world economy tanked, some
new and old ways to demand more say in how some of the sectors thrived, including tech—Amazon founder
world’s wealthiest companies operate. Jeff Bezos would add some $75 billion to his own net
worth in 2020. Back at Bessemer, though, workers
were being pressed to work harder and longer, and
ANDREA DAQUINO

they felt dehumanized. They wanted dignity, not just

higher wages.

The workers’ push to join the
Retail, Wholesale, and Department
Store Union (RWDSU) was always
going to be an uphill battle. Amazon
used its ever-growing wealth to fight
the union’s campaign. Bosses used
social distancing protocols meant
to stop covid-19 transmission as
a pretext to stymie communica-
tion between workers. Employer-
favoring federal and state labor
laws allowed management to run
a scorched-earth anti-union effort:
Amazon hired anti-union consul-
tants, flooded employees with text
messages and signs urging them to
“vote no,” and held “captive audi-
ence” meetings where workers were
required to sit through anti-union
lectures.
The results were painful for sup-
porters: 738 votes for the union,
1,798 against. But even as the votes
were being counted, workers around
the country were agitating. On April
7, self-organized workers under
the mantle of Amazonians United
Chicagoland struck against the
company’s “megacycle,” a grueling
10-hour overnight warehouse shift.
Workers organized in California’s
Inland Empire. Nationwide, hun-
dreds of employees from at least 50
Amazon facilities refused to work
during the pandemic. A group in
Staten Island was moving to start its
own grassroots union. In the Twin
Cities, the Awood Center, a work-
ers’ advocacy group for East African
immigrants, convinced the company
to sit down with workers and come
to an agreement about accommo-
dations for religious observances.
The Bessemer fight, and Amazon
organizing as a whole, reflect a new
groundswell of interest in organizing
among tech workers. But that cur-
rent has also raised a question: What
is a “tech” worker, anyway? The
term could reasonably be applied to
anyone from programmers to data
Review
center staff to warehouse pickers
to assembly-line autoworkers in a
Tesla factory.
The reality is that organizing
in “tech” is sort of like organizing
in “industry” in the 1930s. Back
then, the Congress of Industrial
Organizations shifted the focus of
the labor movement from organiz-
ing the skilled trades to bringing
together the “unskilled” workers
in massive new factories. The new
labor movement was epitomized by
Detroit’s auto plants but emerged
across a wide variety of industries
characterized by new technology
and scientific management tactics.
This era of industry required a new
kind of union, and labor struggled
for decades before hitting upon
methods that worked—and, impor-
tantly, getting the backing of the
Depression-era federal government.
Organizing tech workers will
require a similar effort, a similar
reorganization of labor tactics, and,
quite possibly, a similarly supportive
federal government. As the result in
Bessemer shows, today’s workers
are up against the world’s richest
companies—companies with the
world’s most sophisticated surveil-
lance and information systems, not
to mention millions to spend on
anti-union consultants. For both
sides in this struggle, though, the
bottom line is not money but power.
SHARED POWER
To understand tech’s new labor
movement, says Emma Kinema, an
organizer with the Communications
Workers of America’s tech-sector
organizing project CODE-CWA,
one must understand that tech is
everywhere, that most workers are
in some sense working with technol-
ogy, and that nevertheless there is a
consistency to the thing we call the
“tech industry,” even if it is massive
and diverse.
75
It’s also important to remember
that the culture of Silicon Valley
was anti-union from the start; one
reason California became the tech
hub of choice was that the Boston
area, where many early leaders in
the industry got their start, had a
long-established union presence.
Logic Magazine’s Ben Tarnoff notes
that the common perks and ame-
nities of the tech workplace—free
food, toys and games, and casual
dress—began as explicitly anti-
union measures. This culture, which
sociologist Andrew Ross has called
“no-collar,” was designed to engen-
der not just loyalty but a love for and
identification with the company.
Labor historian Nelson
Lichtenstein says Big Tech has a ten-
dency to lean on its transformational
image to paper over any labor com-
plaints and minimize them as quib-
bles that are impeding the evolution
of the world. That’s nothing new, he
argues—Henry Ford responded in
much the same way when workers
in his factories spoke up, and Ford
learned the tactic from the railroad
magnates who preceded him.
Despite the bosses’ pushback,
workers in these trailblazing sectors
did eventually unionize—though
it took years and quite a few failed
efforts. Steelworkers held massive
strikes in 1919 but failed. “In the
’20s and ’30s everyone thought steel
was impossible to organize,” says
Kinema. “Those workers were too
well paid; their industry was too
new; they had these new, modern
ways of management.” But by 1937,
US Steel recognized the union. “Can
you imagine if Google workers went
on strike?” she asks. The site reli-
ability engineers alone, who main-
tain the technical infrastructure of
Google, “could shut down half the
internet.”
Chewy Shaw is one of those
engineers, as well as the executive
76
P O W E R P L AY
vice chair of the Alphabet Workers
Union, which is part of CWA. The
union went public in January of
2021 with just over 200 members
and now has more than 800, includ-
ing Google programming staff like
Shaw and executive chair Parul Koul,
as well as researchers, data-center
workers, temps, and vendors.
The union is not big enough to
level the internet anytime soon,
nor is a strike imminent (Alphabet,
Google’s parent company, employs
135,000 people). But it has chal-
lenged the company to do better
by its workers, leveraging its com-
plement of “no-collar” staff like
Shaw and Koul to win real changes
for employees with less power
and security. As one example, the
union supported Shannon Wait, a
data-center technician employed
through a subcontractor in South
Carolina, through a wrongful sus-
pension this March for speaking to
coworkers about her working condi-
tions. The National Labor Relations
Board reversed the suspension and
ordered the company to post notices
informing workers that they have
the right to organize.
Much of this work builds on previ-
ous tech-worker activism that gained
steam during Donald Trump’s presi-
dency, when liberal-leaning no-collar
employees learned that the bosses
they had thought shared their val-
ues were in fact happy to work with
the administration. When Google
workers realized they were building
infrastructure for Project Maven, an
artificial-intelligence project for the
US military, they noted that program-
mers working on the software might
not even know that their code could
be used for drone attacks.
Thousands of Google employ-
ees signed a letter protesting the
company’s involvement in Project
Maven in early 2018, and Google
let its Maven contract expire the
Change
next year. That wasn’t workers’ only
THE GOOGLE concern, though—a massive inter-
WALKOUT national work stoppage followed in
UNDERSCORED November 2018, centering on sex-
ual harassment and discrimination
THE FACT at the company.
THAT MANY The Google walkout underscored
WORKERS, EVEN the fact that many workers, even
those with the highest salaries or the
THOSE WITH most job security, didn’t feel valued
THE HIGHEST by the company. And while many
SALARIES OR observers scoffed—venture capi-
talist Mike Solana wrote on Twitter
THE MOST JOB that workers like Shaw and Koul
SECURITY, “aren’t oppressed coal miners”—this
DIDN’T FEEL feeling led to the formation of the
Alphabet Workers Union. “Asking
VALUED BY THE for respect on the job is not specific
COMPANY. to coal miners, and that’s really why
we all do this,” Koul says.
These workers want to leverage
the power they have within the com-
pany as part of a broader working-
class movement. That meant refusing
to work on Maven; and now it means
demanding, in solidarity with the
Black Lives Matter movement, that
the company not sell tech to police.
This organizing takes cues from
unions like the Chicago Teachers
Union, which made racial justice
and economic inequality across the
city central to its demands—and
won those fights through two widely
publicized strikes in 2012 and 2019.
Longtime union strategist Stephen
Lerner says that through such “bar-
gaining for the common good” at
Google, the workers challenge the
company’s impact on society, not
just their own treatment. “I don’t
think the tech organizing would
have any of the kind of resonance it
has now if people opened up with
‘Well, we need a better 401(k) plan,’”
Lerner says.
THE RISE OF THE TECH WORKER
Even in the early 1990s, when
Lerner went to war with Apple as
an organizer of the Justice for Janitors
campaign and won union rights for
subcontracted cleaning workers
across the tech sector, the question
of “Who is a tech worker?” loomed
large. Through those successful cam-
paigns, Lerner helped extend the
definition of a tech worker to virtu-
ally anyone who makes a tech com-
pany run. Cori Crider, an attorney
with Foxglove, a firm that aims to
challenge the power of Big Tech, has
been working with subcontracted
content moderators—real humans
who sift through posts with violence
and racism and graphic sex every
day, trying to determine what vio-
lates a constantly shifting set of rules.
Those workers are often bound
by nondisclosure agreements that
keep them from speaking publicly
about their working conditions. That
allows companies like Facebook to
deny they exist—an assertion the
company stuck with last year even
after reports emerged that modera-
tors working for the outsourcing firm
Accenture were being pushed back
into the office during the pandemic.
Tech workers outside the normal
definition of “employees” are still
finding ways to organize and protect
themselves. Coworker.org, a cam-
paign platform for labor organizing,
is using donations from well-off tech
workers to build a “solidarity fund”
distributed to workers on the other
side of the tech supply chain. Gig
workers on Amazon’s Mechanical
Turk platform are using the site
Turkopticon to come together and
fight for better terms.
At the other end of the tech-
worker spectrum are those build-
ing electric cars at Tesla’s plant
in Fremont, California. Before
Elon Musk’s company bought the
Fremont facility, it was known as
New United Motors Manufacturing,
Inc., or NUMMI, a collaboration
between General Motors and Toyota
where Japanese “lean production”

was brought to America. NUMMI
didn’t survive GM’s bankruptcy in
2008, and Tesla snatched it up.
Cooperating with the United
Auto Workers was one of NUMMI’s
big innovations, but Tesla’s gone
another way. Recently, an admin-
istrative judge at the NLRB ruled
that several of the company’s actions
in response to worker organizing
were illegal—including a couple of
Musk’s tweets as well as harassment
of workers passing out union pam-
phlets, banning of pro-union T-shirts
and buttons, and the interrogation
of organizers and firing of one. The
NLRB’s penalties amount to little
more than a finger-wag—Musk must
read a statement telling workers that
they have the right to unionize, and
rehire the fired worker. He’s appealed
the decision anyway.
The workers at the plant, even
the union supporters, are enthusi-
astic about producing electric vehi-
cles, but they note that the technical
sophistication of the plant does not
prevent a lot of backbreaking manual
labor—or injuries. Jose Moran, one
of the leaders of the union drive and
a former NUMMI worker, wrote a
blog post about the things he wanted
to improve, including the grueling
pace of the work and some badly
designed machinery.
Autoworkers have struggled with
machinery since the days of Henry
Ford. But Tesla workers’ stories echo
the complaints of autoworkers in the
1960s who battled “speed-up”—the
way management would use new
technology to ratchet up the pace
of work—in places like Lordstown,
Ohio, and Detroit. A wave of rebel-
lions within the unions and wildcat
strikes challenged the idea that auto-
mation was making their jobs easier.
As machines sped up the manu-
facturing process, workers had to hus-
tle faster to keep up. The autoworkers
at Tesla, far from representing a labor
Review
aristocracy among autoworkers, say
they make less than unionized work-
ers at GM and Ford. As Moran wrote,
“I often feel like I am working for a
company of the future under work-
ing conditions of the past.”
THE LONG GAME
In the Amazon warehouses, too,
everything old is new again. “The
auto industry tried to do lots of auto-
mation back in the ’80s, ’70s, what-
ever, and they basically plateaued out
where they couldn’t do it anymore.
And Tesla basically tried to do the
same thing,” says Tyler Hamilton,
an Amazon warehouse worker from
Minneapolis. “It’s the same thing
with Amazon. There’s only so much
you can do with automation.”
Mohamed Mire, a coworker of
Hamilton’s, explains that most of
Amazon’s vaunted technology goes to
tracking the workers rather than mak-
ing the work efficient. Scanners that
the workers use to scan packages also
keep track of their so-called “time off
task,” and they get written up if their
productivity rate falls. Robots that
Hamilton likens to “giant Roombas”
carry merchandise around the ware-
house but malfunction often—lately
his job has included setting the robots
right when they stop working. Data
from Amazon shows that injury rates
are higher at facilities with robots
than without them.
Hamilton and Mire work with the
Awood Center, which—since it is a
worker center rather than a union—
doesn’t go through NLRB elections
but instead organizes through direct
action. Awood members have won
some concessions from Amazon,
particularly around prayer time
(many are practicing Muslims) and
accommodations for fasting during
Ramadan. They’ve also gotten people
hired back who’d been fired.
Despite the results in Alabama,
workers like Hamilton and Mire have
77
no intention of slowing their orga-
nizing. But Amazon’s heavy-handed
tactics—including the hiring of actual
Pinkertons, security agents from
a company that has been helping
employers break unions since the
19th century—are also unlikely to
stop. The NLRB is deciding whether
to consolidate complaints against the
company across its various regions—
there have been at least 37 in 20
US cities since the beginning of the
pandemic. RWDSU filed 23 com-
plaints of unfair labor practices just in
Bessemer, including the charge that
Amazon illegally threatened workers
with layoffs or the facility’s closure.
There is clearly still a long way
to go before tech workers win at
the bargaining table, but history
offers them plenty of models to
look to. Lichtenstein, the labor his-
torian, points to the International
Longshore and Warehouse Union,
a powerful West Coast waterfront
union that inked an agreement with
shippers in 1958 to get dockworkers
a slice of the gains from automation.
When selective use of automation
led to more injuries, the union actu-
ally pushed for more tech to improve
safety. They drove wages for what
had been insecure, contingent work
to over $150,000 a year.
In the current struggle, the Biden
administration has signaled sup-
port both for sweeping labor law
reform—which would make many
of Amazon’s tactics at Bessemer
illegal—and potentially for regulat-
ing Big Tech.
And Hamilton notes, “It took
something like 50 years to unionize
US Steel. Amazon’s warehouses were
just built a handful of years ago. If
it’s not this year or next year, it’ll be
five years from now.” Q
Sarah Jaffe is a fellow at Type
Media Center and author of Work
Won’t Love You Back.
78 Change

THE

JAMES SUROWIECKI
or Apple, Amazon, Facebook, and Alphabet,
F covid-19 was an economic blessing. Even as the
The big breakup pandemic sent the global economy into a deep
recession and cratered most companies’ profits, these
It’s become trendy to talk of forcing the biggest US tech companies—often referred to as the “Big Four” of tech-
companies to loosen their monopolistic hold on the digital nology—not only survived but thrived. Collectively, they
economy. But it won’t be easy. now have annual revenue of well over a trillion dollars,
and the value of their stocks has soared: together they’re
worth $2.5 trillion more than they were 15 months ago.
ANDREA DAQUINO

Yet at the same time, they have come under unprec-

edented attack from politicians and government

regulators in the US and Europe.
While congressional hearings on
charges that Facebook has been
censoring conservatives or not
doing enough to restrain disinfor-
mation and hate speech may have
gotten most of the headlines and
public attention, the companies are
facing far more substantive threats,
in the form of new lawsuits, pro-
posed bills, and regulations.
This past fall, the Federal Trade
Commission and 48 state attor-
neys general filed suit against
Facebook, charging it with illegally
maintaining a monopoly over the
social-networking space “through
a years-long course of anticompet-
itive conduct.” Soon after, the US
Department of Justice and 11 state
attorneys general filed suit against
Google, charging it with illegally
maintaining a monopoly over the
search and search advertising mar-
kets. Apple is currently locked in
a civil trial with game developer
Epic Games, which is challenging
Apple’s control of its App Store on
antitrust grounds.
Last summer, the US House
Judiciary Committee concluded a
19-month investigation into alleged
anticompetitive activity by the tech
titans. The resulting 450-page
report described the companies
as “the kinds of monopolies we last
saw in the era of oil barons and rail-
road tycoons” and recommended
that the government take action
against them.
It’s easy, of course, to dis-
miss anything that comes out of
Washington or Brussels as politi-
cal posturing, but in this case that
would be a mistake. President Joe
Biden has named some of Big Tech’s
sharpest and most vocal critics—
including Columbia University
professor Tim Wu, author of the
book The Curse of Bigness, and
Lina Khan, who served as special
Review
counsel to the Judiciary Committee
during its investigation—to import-
ant roles in his administration.
Europe is putting in place tougher
regulations to try to limit Big Tech’s
power. And antitrust action, at least
with regard to the tech industry,
has become that rarest of things: a
bipartisan issue in Congress.
What’s arguably more import-
ant is that we’re in the middle of
a radical shift in the intellectual
discussion—one that has made it
much easier to go after Big Tech.
In many ways, we seem to be going
back to the antitrust vision that
determined US policy toward big
companies for much of the 20th
century, a vision that’s much more
skeptical of the virtues of size and
much more willing to be aggressive
in keeping companies from exer-
cising monopoly power.
America’s key antitrust laws were
written around the turn of the 20th
century. The Sherman Antitrust Act
of 1890 and the Clayton Act of 1914
remain on the books today. They
were written in broad, far-reaching
(and ill-defined) language, targeting
monopolists who engaged in what
they called “restraint of trade.” And
they were driven in large part by
the desire to curb the giant trusts
that had, via a series of mergers
and acquisitions, come to domi-
nate America’s industrial economy.
The quintessential example was
Standard Oil, which had built an
empire that gave it essentially com-
plete control over the oil business
in the US. But antitrust law wasn’t
just used to block mergers. It was
also used to stop a host of practices
that were deemed anticompetitive,
including some that nowadays seem
routine, like aggressive discounting
or tying the purchase of one good
to the purchase of another.
This all changed with the Reagan
administration in the 1980s. Instead
79
of worrying about big companies’
impact on competitors or suppli-
ers, regulators and courts started
to focus almost entirely on what
was called “consumer welfare.”
If a merger, or a company’s prac-
tices, could be shown to lead to
higher prices, then it made sense
to step in. If it didn’t, antitrust reg-
ulators generally took a hands-off
approach. That’s why Facebook’s
acquisitions of Instagram and
WhatsApp, Amazon’s acquisition
of Zappos, and Google’s acquisitions
of DoubleClick, YouTube, Waze, and
ITA all sailed through the regulatory
approval process without a hitch.
No longer, though. Over the past
four or five years, scholars, politi-
cians, and public advocates have
begun to push a new idea of what
antitrust policy should be, arguing
that we need to move away from
that narrow focus on consumer wel-
fare—which in practice has usually
meant a focus on prices—toward
consideration of a much wider range
of possible harms from companies’
exercise of market power: damage to
suppliers, workers, competitors, cus-
tomer choice, and even the political
system as a whole. They’ve done so,
not surprisingly, with the Big Four
squarely in mind.
But what exactly would reining
in Big Tech’s power look like? Short
answer: It depends very much on
which company you’re going after.
THE TARGETS
While antitrust advocates often
rhetorically lump Apple, Amazon,
Google, and Facebook together,
creating a memorable image of four
giant “gatekeepers” collectively
controlling access to the digital
economy, in reality the four compa-
nies have very different businesses
that raise very different antitrust
questions and will lend themselves
to very different antitrust solutions.
80
THE BIG BREAKUP
Take, for a start, Apple. It is
the most valuable company in the
world, as of this writing worth
more than $2 trillion. It’s also the
most profitable company in the
world. And yet, when it comes to
discussions of antitrust and Big
Tech, Apple often seems like an
afterthought. In Wu’s book, Apple
barely makes an appearance, and
in Senator Amy Klobuchar’s new
book, Antitrust, which is a ringing
call for remaking and enforcing
anti-monopolization policy, the
discussions of Apple seem more
cursory than central to her thesis.
That may be in large part
because Apple has become a behe-
moth mostly on its own—while it
has made plenty of acquisitions,
its recent growth is mainly due to
the simple fact that it has intro-
duced three of the most successful
and lucrative technology products
in history, and that it has contin-
ued to convince customers to keep
upgrading to the next generation of
products. Even in this new world,
it is not illegal to become hugely
successful by building the prover-
bial better mousetrap.
To be sure, Apple has antitrust
issues, which center on its require-
ment that all developers who are
making apps for the iPhone and
iPad sell their goods through the
App Store, with Apple collecting a
30% fee. So it’s possible Apple will
end up having to let developers sell
directly to consumers, or even allow
independent app stores. Even so,
it could still collect a licensing fee
from any app that wanted to be on
the iPhone. And most users would,
in all likelihood, continue to use the
App Store regardless, if only out of
habit and convenience.
So in the grand scheme of
things, Apple wouldn’t seem to
have that much to worry about from
increasing antitrust pressures.
Change
Amazon’s situation is more
complicated. It, too, has the fact
of organic growth going for it; while
it has made its share of acquisitions,
it has grown mostly on its own,
driven by its relentless appetite for
The Curse of
selling more, its huge investment in
Bigness
infrastructure, and its willingness
Antitrust in the
to spend huge amounts of money
New Gilded Age
in order to win and keep customers.
Its biggest antitrust problem stems,
paradoxically, from something it
created itself: Amazon Marketplace.
Marketplace was the result
of a decision that, at the time,
seemed crazy to many: allowing
outside sellers to compete with
Amazon products and sell on its
platform, with Amazon taking a
Antitrust cut of the proceeds. It turned out
to be a genius move: Marketplace
Taking On
Monopoly now accounts for a huge chunk of
Power from the Amazon’s sales and an even big-
Gilded Age to ger chunk of its retail profits. But
the Digital Age
Marketplace has also become the
place where Amazon’s exercise of
power is most visible and most
obviously problematic.
As Brad Stone details in his
new book Amazon Unbound, many
Marketplace sellers accuse the com-
pany of juking search results to
reward those who use its fulfillment
services rather than filling orders
on their own; rewarding sellers
who advertise on the site; boost-
ing Amazon’s own house-brand
products in the rankings; and, most
famously, using Marketplace data
to identify particularly successful
products and then mimicking them
to undercut Marketplace sellers.
Whether Amazon is a retail monop-
olist is an open question—its total
sales remain well below Walmart’s,
and even in online commerce its
market share is below 50%. But
it does unquestionably control
Marketplace, and the sellers who
use it don’t have many other places
to go. That’s why politicians like
Senator Elizabeth Warren have
argued that Amazon should be
required to spin off Marketplace,
while others have suggested that
tough regulations be imposed on
how it manages the site.
Even so, it’s not surprising that
when the government was decid-
ing which companies to file anti-
trust lawsuits against, it went after
Google and Facebook first. Those
companies are the easiest to fit into
a traditional definition of a monop-
oly—more than 90% of all internet
searches are performed through
Google, and it and Facebook
together control around 80% of the
digital ad market. Google’s acquisi-
tions of DoubleClick and ITA played
key roles in fueling its evolution. It
faces a lawsuit in Europe for tinker-
ing with search results to put its
own shopping-comparison engine
higher up in the rankings and the
sites for rival services lower down.
Perhaps most important, Google
effectively holds the economic fate
of websites across the world in
its hands—a change to its search
engine or to YouTube’s algorithms
can cost people thousands of cus-
tomers or viewers. None of this
might have mattered much in
the days when regulators wor-
ried mainly about a monopoly’s
impact on consumer prices, since
just about everything Google does
is free to consumers. But under the
new antitrust model, the company’s
sheer reach makes it a good target.
Not, though, as good a target
as Facebook. If you had to bet, in
fact, on which company is most
likely to suffer real consequences
from the revolution in antitrust pol-
icy, you would be smart to bet on
Facebook. It gets 61% of all social
media visits in the US. It’s been
famously ruthless in snuffing out
competitors, either by duplicat-
ing their features—as it did with

Snapchat and Twitter—or by simply
acquiring them. Its acquisitions of
WhatsApp and Instagram look like
precisely the kind of anticompet-
itive acquisitions that regulations
were designed to stop. And its lack
of transparency about the way it
uses customer data has made it
notorious.
BUT WHAT WOULD A BREAKUP
REALLY LOOK LIKE?
The Big Four are unquestionably
in the government’s crosshairs. Yet
their stocks are more valuable than
ever, which suggests that investors,
at least, are betting that the anti-
trust hullaballoo won’t add up to
much. Why?
One reason is that in going after
Big Tech, trustbusters are going
after some of the most popular com-
panies in America. Surveys rou-
tinely find that Amazon is the most
trusted company in the US, with
Google and Apple not far behind
in the “most admired” rankings.
Facebook is the exception; but even
if people don’t like it, they find it
useful.
Antitrust advocates want to take
other kinds of harms into account,
but they’re not saying that con-
sumer interests should be ignored.
And the benefits people get from
these companies are easy to show,
while the harms they’re inflict-
ing on users can be hard, if not
impossible, to define, often rest-
ing on somewhat abstract ideas of
restricted consumer choice and
the costs of lost future innovation.
These costs are arguably real, but
it’s not obvious they’re enough to
build popular support for remedies
like breaking the companies up. And
while in theory we’re talking about
the law, in practice all decisions
about what cases to bring and whom
to bring them against—and which
regulations and remedies politicians
Review
and regulators push for—are shaped
THE BENEFITS by politics, which means in turn that
PEOPLE GET they’re shaped by popular opinion.
FROM THESE It’s unlikely any president is going
to want to be seen as the person
COMPANIES ARE who broke up Google, particularly
EASY TO SHOW, if it means worse search engines
WHILE THE and maps.
What this suggests is that even
HARMS THEY’RE if public rhetoric suggests a cam-
INFLICTING ON paign to cut Big Tech down to size,
USERS CAN BE we’ll likely end up instead with a
series of company-specific rem-
HARD, IF NOT edies. Amazon may have to com-
IMPOSSIBLE, ply with stricter regulations on
TO DEFINE. Marketplace, including curbs to
its power to manipulate its search
results or perhaps even its ability to
compete with Marketplace sellers.
Apple’s monopoly on the App Store
may end. Google may face stricter
regulations on what it can do with
data, and how its search engine’s
ranking works.
These would not be trivial
changes, which is why the compa-
nies can be expected to fight them.
And yet in most cases, it’s hard to
see that they would be transforma-
tive. In fact, in recent years these
companies have already had to
change various questionable prac-
tices in response to court cases or
inquiries from regulators. It hasn’t
kept them from missing a beat.
Facebook, which is the least
popular of the Big Four, might be
different. It may be at risk of the
kind of breakup that happened
to Standard Oil and AT&T, with
Instagram and WhatsApp spun out
as independent companies. That
would be logistically difficult, since
Facebook has worked assiduously
to integrate the three services. But
it’s not impossible. And it’s a log-
ical, easy-to-understand remedy
that might inject some competition
into social media. Even so, it’s not
clear this would fundamentally dent
Facebook’s hold on users, given the
81
treasure trove of data it controls
and the power of network effects.
Indeed, if the new antitrust
movement really wants to change
the digital economy, challenging the
Big Four’s various sketchy practices
is not going to be enough. These
companies’ greatest competitive
advantage isn’t the legally dubi-
ous stuff they’re doing—it’s their
perfectly legal access to enormous
amounts of detailed and granular
user data. That data helps them
understand their users better than
anyone else and make continuous
improvements to their products
and services—which in turn helps
them keep their current users and
add new ones, which gives them
access to more data, and so on. It
is the key to their growth.
Truly challenging the power of
the Big Four would mean rethink-
ing how data is gathered and used
by companies, and who gets access
to it. It might mean requiring that
data be shared, that algorithms be
transparent, and that consumers
have far more control over what
they share and what they don’t.
For that to happen, the new
trustbusters will have to make the
case that even if we like what our
digital overlords are doing with
our data, it’s still wrong for a small
number of companies to control so
much of it. In a way, they need to
make the case that, as in the past,
at some point bigness in and of
itself is a curse. Big Tech has made
that a hard sell in America, simply
because the companies have cre-
ated so much value for consumers.
We’re going to find out if that’s
enough to keep them safe in this
new world. Q
James Surowiecki is the author
of The Wisdom of Crowds and
formerly wrote the Financial
Page for the New Yorker.
82 Change

n late 2007, less than 10

I years into the company’s
existence, Google came out
swinging on the clean energy front.
To a fanfare of plaudits up and down
Silicon Valley and well beyond, it
declared “RE<C” as its goal: make
renewable energy cheaper than coal.
The company invested tens of mil-
lions of dollars into R&D efforts from
concentrated solar power to hydro-
thermal drilling. Four years later,
those efforts had been scrapped.
It would be all too easy to see this
as an admission of failure—big tech
playing in an arena it knew nothing
about, with the hubris that Silicon
Valley is known for. But something
else was going on. Google’s shift
in strategy was a reflection of the
growing success of the solar sector.
Google realized its energies were bet-
ter directed toward massively scaling
up existing renewable technologies
that had plummeted in price, rather
than inventing new ones.
While Google nailed the switch
from R&D to deployment, it argu-
ably still bet big on scaling up the
wrong technology. In the early
2010s, the solar race looked like
a tight competition between solar
photovoltaic (PV) and utility-scale
concentrated solar power (CSP),
which uses sun-heated fluids to
drive power turbines. Google quickly
THE invested more than $1 billion in a
slew of renewables companies and
utilities, including big investments
GERNOT WAGNER
in CSP outfits BrightSource Energy
and eSolar. A decade later, such
choices aren’t looking promising,
Cheap solar PV and as CSP, too, has been losing out to
PV’s continuing rapid cost declines.
expensive climate change Google is not alone in repeatedly
misjudging the dropping price of
The rapidly dropping price of solar power has transformed solar cells over the last few decades
how we think about clean energy. and its impact on how we think about
clean energy. Solar PV costs fell
ANDREA DAQUINO

roughly by a factor of 10 in the past

decade, on top of already impressive

cost declines up to that point, for
a total decline of around a factor
of a hundred since US President
Jimmy Carter unveiled solar panels
on the White House in 1979. (Ronald
Reagan took them down in 1986,
during his second term as president.)
To put it in perspective, if gasoline
had similarly dropped in price from
1979 levels, it would cost pennies a
gallon today. Gasoline, of course, is
a commodity, with prices fluctuating
for a number of technological, eco-
nomic, and political reasons. Solar
PV prices are also driven by all these
factors, but over the years, technol-
ogy has clearly dominated. (This year,
prices for solar PV modules have
increased by around 18% because
of a temporary crunch in the silicon
supply chain.)
In its latest annual World Energy
Outlook, the International Energy
Agency declared solar PV to be
“the cheapest source of electricity
in history” for sunny locales with
a low cost of financing. These two
qualifications are important. Sun is
obvious—solar is always going to be
cheaper in Phoenix, Arizona, than in
New York City—but the report con-
cluded that solar is now cheaper than
coal and natural gas in many places.
Financing is key to why this is
true. Solar PV and other renewables
such as wind have low or close-to-
zero operating expenses—upfront
costs have always been the big hur-
dle, and financing has been a big rea-
son why. Thanks in part to various
government policies, solar invest-
ment has become much less risky
over the last decade or so, freeing
up cheap money.
As a result, solar PV deployment
has increased rapidly; it’s now the
fastest-growing source of electricity
globally, and figures to be for some
time to come. It’s starting from a
low base of installed capacity, how-
ever, far behind coal, gas, hydro,
Review
nuclear—even wind, which has been
SOLAR NEEDS TO cheap for longer. And therein lies
BE SO CHEAP IT one of the biggest problems for solar
MAKES SENSE TO PV. It might be the cheapest form of
electricity for many, but that on its
BUILD NEW own doesn’t make the clean-energy
SOLAR CAPACITY transition nearly quick enough.
We need ever further techno-
AND SHUTTER logical advances. Why stop at grid
WORKING COAL parity, the point where it’s as cheap
AND NATURAL- to build and operate solar PV as to
supply electricity via fossil energy
GAS PLANTS sources? Why not 10% cheaper?
STILL MAKING Why not strive to slash costs by
MONEY FOR another factor of 10 within a decade?
Such drops are needed because the
THEIR OWNERS. hallowed grid-parity goal is mislead-
ing—the real question is at what
point utilities will actually abandon
existing coal plants and switch to
solar, rather than merely avoid add-
ing new coal capacity. Solar needs
to be so cheap it makes financial
sense to build new solar capacity and
shutter working coal and gas plants
still making money for their owners.
All that calls for policy to both
push existing solar technology and
support R&D in new technologies.
The entire package includes technol-
ogy research, development, demon-
stration, deployment, and diffusion.
Every step along this chain deserves
direct government support, keeping
in mind that it also gets increasingly
more expensive the further down
the chain one moves.
HOW TO GET CHEAPER
To better optimize investments to get
to even cheaper solar, it’s worthwhile
to understand what factors have
driven down the cost of renewable
power over the last few decades.
MIT energy systems scientist
Jessika Trancik and her group find
that the dramatic cost declines in
solar cells over the course of three
decades can largely be attributed
to three factors: R&D leading
directly to improvements in module
83
efficiency (how much of the sun-
light is converted into electricity)
and other fundamental technolog-
ical advances; economies of scale
attributed to the size of solar-cell
manufacturing plants and the
increasing volume of inputs such as
silicon; and improvements achieved
through learning by doing.
None of that is too surprising,
but what is less obvious is that the
relative contribution of each varies
greatly over time. From 1980 to 2000,
R&D accounted for around 60% of
cost declines, with economies of
scale coming in at 20%, and learning
by doing a distant third at around 5%;
other largely unattributable factors
account for the balance. That makes
sense; it was a period of impres-
sive advances in the efficiencies of
solar cells but not a time of signif-
icant manufacturing and deploy-
ment. Since then, the pendulum has
swung from R&D and fundamental
technological improvements toward
economies of scale in manufacturing,
now accounting for over 40% of cost
declines. It’s worth noting, however,
that research advances still account
for some 40% of declines.
The lesson for future invest-
ments that aim to make solar even
cheaper: there should be direct sup-
port for all three, skewed toward
economies-of-scale factors. Trancik’s
findings only consider the solar PV
module itself. That still leaves instal-
lation, connection to the grid, and
other factors that make up total
system costs. These are areas that
will likely be improved as techni-
cians and companies become more
experienced. While the results of
subsidies for increasing solar PV
installations appear to be mixed at
best, policies such as feed-in tariffs,
which offer favorable long-term
contracts to solar PV producers, and
renewable portfolio or clean energy
standards, which set quantity targets
84 Change
CHEAP SOLAR
for renewables, show clear results in
driving overall deployment.
NO FREE LUNCH
Despite the dropping price of solar,
the transition to renewables will
still be costly. The big question, of
course, is how expensive compared
with what—climate change, too,
comes with costs. Cheap solar gets
even more financially attractive to
But developers if the social and environ-
mental costs of carbon emissions
from fossil fuels are considered.
wait, A lot here hinges on the social
cost of carbon (SCC), a tally of the
financial damage each metric ton of
carbon dioxide emitted today causes
there’s to the economy, society, and the
environment—and, by extension,
how much each ton of CO2 emitted
more. should cost. It’s a number that says
a lot about the true cost of coal and
other fossil fuels—and about the
appropriate support for solar PV
and other renewables.
The latest US SCC, calculated by
Your subscription the Biden administration, puts the
gets you more than number at around $50 for a ton of
just this magazine CO2 emitted now. But that is surely
an underestimate. Some calculate
the SCC to go over $300 per ton of
CO2, after fully accounting for the
You’re already a future damage caused by carbon
subscriber. Register emissions and for the uncertainties
your account and about climate change.
start enjoying: Whichever number you settle on,
it means coal, oil, and natural gas will
• Unlimited web access be far more expensive if you account
for the full costs of greenhouse-gas
• Exclusive digital emissions. Only then will low-carbon
stories technologies be on the same playing
• The Algorithm field as fossil fuels.
An explicit carbon price via a
newsletter
tax or emissions trading system
• Subscriber-only app should be among those steps—
but it ought not stop there. For one
thing, renewable portfolio or clean
energy standards, too, establish a
price on carbon. Current US renew-
technologyreview.com/ able portfolio standards at the state
subonly
level translate into equivalent carbon
prices of around $60 to $300 per
ton of CO2, well within the range of
recent SCC estimates. A federal clean
electricity standard, part of the Biden
administration’s proposed American
Jobs Plan, could be in a similar range
and would be similarly justifiable on
the basis of updated SCC ranges.
Such a federal clean electric-
ity standard would be a real boon
to solar PV and other renewables,
but climate policy must not end
with pricing carbon. It also needs to
include direct subsidies for deploy-
ment and support for R&D.
The most productive policy
sequence might go something like
this: first drive down the cost of
renewables to create an econom-
ically viable alternative to high-
carbon fuels, then price carbon via
a direct price, a clean electricity
standard, or something similar. The
combination of the two should then
lead to rapidly deployment of renew-
ables at scale. In many ways, that is
precisely what has happened, and it
points to the clear need for the Biden
administration and others to push
for a price on carbon in whichever
form that might happen.
But if far cheaper solar PV is the
goal, it will also be critical to increase
R&D to drive further improvements
in the efficiencies of solar cells and
find manufacturing advances that
will allow even greater savings. And
it is essential to keep exploring the
scientific frontiers in search of other
solar materials that could one day
be even more efficient and cheaper.
Solar PV is cheap, but it is not
free. Paying the price to make it ever
cheaper will be well worth the cost. Q
Gernot Wagner teaches
climate economics at New
York University. He is the
author of the forthcoming
Geoengineering: The Gamble.
 Review 85

In 2017, MIT Technology Review

honored Tracy Chou as one of our
35 Innovators Under 35. At the time,
Chou was working to expose Silicon
Valley’s diversity issues. As an engi-
neer at Pinterest, she’d published a
widely circulated blog post calling
for tech companies to share data on
how many women worked on their
engineering teams. She collected their
responses in a public database that
revealed how homogeneous many
technical teams at top companies
still were.
About a year later, Chou started
a company called Block Party that
targets online harassment by giv-
ing Twitter users more control over
which tweets appear in their feed
and mentions. The service signed up
its first paying customers in January.
With her new company, Chou
wants to fix some of the problems
she’s experienced firsthand in the
tech industry—including the sort
of online harassment of which she
has been a target. Here, we check
in with Chou, who is based in San
Francisco, to learn more about what
it takes to make change in the tech
sector and what entrepreneurs like
her are up against.

hen we last spoke, I

THE W had just left Pinterest.
I’ve always been drawn
to smaller companies: I joined
WUDAN YAN
Pinterest when it had about 10
employees and left when it had
about 1,000. It felt like time for me
Roadblocks to move on and do something new.
I’ve worked for so many startups
An entrepreneur shares what happened when she set and have come to recognize some of
out to try to solve some of the problems she experienced the structural issues around start-
firsthand in the tech industry. ups and funding and how those
factors influence what problems
ANDREA DAQUINO

get solved. A lot of founders natu-

rally work on problems that directly
affect them: it’s easier to know
86
R OA D B LO C KS
what’s important or what could
be improved by technology.
In thinking about my next
steps, I thought about products
I’ve worked on and checked that
against questions like, Do I care
about this? Is there something to be
made that can be commercially via-
ble? There are lots of really import-
ant issues that will not be solved
naturally through a startup.
I ended up on Block Party,
which pulls together a few differ-
ent threads from my background.
I’ve worked as an engineer at vari-
ous social platform companies, and
I’ve worked on monitoring, moder-
ating, and increasing the quality of
content, and figuring out how prod-
uct design influences community
behavior. Not only did I build mod-
eration tools at Quora that reviewed
content quality, but I also took puni-
tive actions against people who
violated the site’s policies.
I’d also spent a lot of time look-
ing at how the lack of diversity and
representation in teams meant that
products were built in a skewed
way. For instance, nondiverse teams
of people who generally don’t get
targeted with abuse and harass-
ment don’t tend to build protections
against that in their apps.
The last part of my background
that led me to Block Party was
just getting targeted more with
harassment. Over the last year, I’ve
definitely gotten more anti-Asian
harassment online. Some of it was
truly targeted at me by individuals,
and other times I would attract
trolls just by having a presence
online.
I got online very young, and at
first the internet was a fun way
to connect with friends. I was on
AOL Instant Messenger, which
was a better way to chat with my
friends in high school: I didn’t have
a cell phone, and I couldn’t hog the
Change
phone line that I shared with my
family. I was also on some of the
blogging platforms, like Xanga and
LiveJournal. They were nice outlets
at the time.
Pretty early on, though, some-
Tracy Chou on one set up an anonymous Xanga
the cover of our page dedicated to hating me. I
35 Innovators
Under 35 issue
think it was someone from school,
in 2017. because it referenced things from
high school. A lot of it was hat-
ing on me because I did well aca-
demically. It didn’t bother me as
much at the time as it did when I
got older and looked back on it.
Back then, I thought this person
was just insecure and jealous. I
thought it was a bit sad and messed
up that someone would write full
posts dedicated to trying to take
me down.
I didn’t report it. Who would I
have reported it to? It didn’t even
cross my mind to go to my school
and report it. And I didn’t neces-
sarily want my teachers or school
administrators to see the page
either, since it was pretty hateful
content.
My parents didn’t raise me to
be someone who was outspoken
and challenged the status quo. I
definitely wasn’t encouraged to
speak up against the system in
any way. Like many other children
of Asian immigrants to the US, I
grew up believing that this is not
my country, and my parents and I
are here trying to find opportuni-
ties for ourselves. We didn’t have
a safety net. I grew up more with
a head-down mentality of do good
work, work hard, and try to make it.
My dad, who’s an engineer, gave
me a philosophical thought exper-
iment when I was quite young: If
you could be reborn as anybody in
the world tomorrow, how would
you design the world today? You
wouldn’t want to design a world
that’s vastly unequal, where most
people are at the bottom, because
that could very likely be you if you
were born as anybody tomorrow.
You’d want to design a much more
equal world. That got me thinking
I didn’t like that the world was so
unequal and so many people were
much less lucky than me.
That feeling has made me take
the privilege that I have and pay it
forward to make the world a little
more just. I went to Stanford; I’ve
worked at companies that people
within tech find credible. So I can
try to amplify more voices or dif-
ferent perspectives.
When I decided to pursue Block
Party, I did a bit of research and
talked to others who have worked
in anti-harassment, and to people
who were building solutions, to
understand what the market was
like. Not a lot of companies under-
stood the true user experience of
dealing with harassment—it felt
like many people were approach-
ing this problem purely from a
business angle, because they didn’t
know what it was like to experience
it themselves. Some people were
building machine-learning models
to detect toxic content and thought
that would solve the problem.
As someone who deals with
this, I understand the emotional
impact much more. There are
certain things that are emotion-
ally distressing to read even if
they’re unlikely to be flagged by
an algorithm. For example, I once
received a long message from a
man which essentially said, “You
should smile more and then you’ll
be more attractive.”
That message bothered me
for a few weeks before I realized
how gaslighty and inappropri-
ate it was. The impact was very
emotional, and the meaning of
the message—not to mention the
anti-feminist and regressive point

of view within—goes deeper than
what would be picked up by any
algorithms.
My challenges in getting fund-
ing for Block Party started when I
applied to be in the winter 2020
class for Y Combinator, an accel-
erator dedicated to funding early-
stage startups. In Silicon Valley, if
you’re looking for a credential, Y
Combinator is one of the better
ones you can get. I applied, iron-
ically, because I anticipated that
YC—which is made up of mostly
white men who wouldn’t neces-
sarily understand the problems I
wanted to take on—would be very
skeptical. My thinking was that if
I had a YC stamp of approval, that
could dispel some of the skepticism
around my product.
I went to the YC campus in
Mountain View, California, for my
interview. They put applicants in
different rooms with a panel of
interviewers, and each room had
a list of the interviewers printed
outside. When I went to my room, I
saw that list of names and thought,
“This isn’t going to be good.” Of the
four names, three were white men.
The interviews were only 10
minutes long, and my panel started
by asking what I was building. I
said, “Block Party is building con-
sumer tools for dealing with online
harassment and gives people more
control over their online experi-
ence.” And the question I got was
“You said this is a consumer prod-
uct, but consumer implies mass
market, and this is a very niche
problem. Right?”
The rest of the interview was
pretty much like that: a strong asser-
tion followed with “Right?” The
whole tone of the interview was:
online harassment was a small prob-
lem that’s already being solved, so
why are you working on this? I came
out of the interview livid.
Review
After YC rejected me, I raised
WITHIN 10 some follow-on funding to extend
OR 15 MINUTES my pre-seed funding, about $1.5
OF THE AMA million. That might feel like a good
chunk of money in some markets,
STARTING, but it’s not enough to hire engi-
TROLLS neers, which means I had to be a lot
TOOK OVER more creative with how I built the
product—I had to build it myself.
THE THREAD, There’s no chance Block Party
PRESENTING would have even gotten to where
BAD-FAITH it is now if I were not myself a very
strong engineer.
ARGUMENTS Last summer, as part of Mozilla’s
AND BURYING Fix the Internet incubator, I was
ALL MY asked to do a Reddit AMA in an
effort to grow Block Party. I ini-
ANSWERS. tially pushed back on that because
my first reaction was: Reddit is the
troll farm. At some point, I talked
myself out of my initial instinct. I
thought, maybe it won’t be so bad!
I’ve seen some good AMAs!
But within 10 or 15 minutes of the
AMA starting, trolls took over the
thread, presenting bad-faith argu-
ments and burying all my answers.
They were like, “Look, she’s too
afraid to answer the questions.”
I had answered, but my answers
would just get buried or disappear.
By the time the Reddit moderators
locked the thread many, many hours
later, there were thousands of com-
ments, the majority of which were
awful, terrible things. It was a really
traumatic experience.
Then there were a couple of
threads on 4Chan that referenced
the Reddit AMA, which led 4Chan
and Reddit trolls to launch an attack
on Block Party. They submitted a
bunch of applications for our waitlist
through a sign-up form and created
hundreds of accounts with my name
and profile photo on Substack and
posted racist, misogynistic con-
tent. For a couple of weeks, it was
a nightmare everywhere I went
online. I couldn’t even log back
on to Reddit because my Reddit
87
inbox was overflowing with harass-
ment. It was ironic that if I build
anti-harassment software, I just get
harassed for it.
I have a stalker—an online
harasser who has moved into real-
world stalking—which made me
more fearful. I wondered, will
these trolls dox me or do some-
thing worse? When I’ve gone to
file police reports on my stalker,
I’ve felt gaslit. The police make
notes like “Victim believes suspect
is obsessed with her,” which under-
cuts the truth: I get thousands of
tweets from this person. He would
fly to where I was. He has entry and
exit records at the airport nearest
to me, and told me what motel he
stayed at when he was trying to find
me in San Francisco.
Someone who works in private
security once told me one of the best
things you can do in that situation is
to flip your mentality into one that’s
very proactive: think, if you were
the stalker, what would you do?
That helps you frame your defenses.
Shifting your mentality to being
proactive means you have agency
and it’s really helpful. Because if you
feel like you’re helpless and under
attack, that’s very debilitating.
As unpleasant as my experiences
have been, I treat this all as user
research for Block Party.
I want online harassment gone,
and if I can be a part of solving this
problem—whether that’s with Block
Party or not—I’m happy. I’m taking
all the experiences that I’ve had in
product engineering, working with
Silicon Valley companies, working
on platform companies, and my own
experiences with harassment and
hopefully turning it into something
positive. Q
Wudan Yan is an independent
journalist based in Seattle,
Washington.
88 The back page

The change
Throughout the decades, change
has proved difficult—whether
you’re trying to make it happen or
just reacting to it.

chronicles

OCTOBER 1967 FEBRUARY 1971 JULY 1994

From “How I Stopped Worrying and Learned
to Live with Irrationality”: Whatever form it
takes, resistance to any change is essentially
an act of self-protection. When an individual
perceives that the consequences of a change
conflict with his needs and desires, he will
become anxious and fear for his future. …
Whether or not the circumstances justify
such response is irrelevant. What is import-
ant is his negative attitude toward the cause
of his anxieties, fears or outrage: the change
and/or its perpetrators. Such a process has
little connection with intellect, logic and
rational thinking. Nevertheless, once one
commits himself to a course of resistance,
he can easily justify his actions to himself by
finding convenient rationalizations. On the
other side of the fence, any resistance to a
change may appear to those responsible for
the change both mystifying and frustrating.
From “Implementing Programs of Social
and Technological Change”: The programs
to which we have committed ourselves,
for example, those based on problems of
poverty, crime, inadequate health care, or
poor housing—have clearly failed to achieve
significant change. … Perhaps we simply
had the wrong ideas. Perhaps we never ade-
quately committed ourselves. But a more
fundamental criticism will serve us better.
The fact of the matter is that we are simply
unable to bring broad programs of social
change into effect. We do not understand
the process, although we have myths about
it, and our failure to understand it preju-
dices our present efforts with respect to
such ideas in good currency as pollution
and environmental control, guaranteed
annual income, and administrative decen-
tralization of social programs.
From “Travels on the Net”: There has been
little discussion of exactly how the new
media will affect our society. This is
frightening because new forms of com-
munication have historically changed the
social fabric in profound and nonobvious
ways. … With only the 94 ASCII charac-
ters that can be produced by typing on a
standard computer keyboard, users con-
struct a parallel world that is sometimes
more compelling than our own. In these
virtual worlds, identity becomes mallea-
ble: our gender is what we say it is. This
not only changes how we interact (by
eliminating the social cues we are accus-
tomed to) but also destroys the notion that
identity is rooted in the body. The ability
to collectively create a world out of words
stretches the imagination and exercises
our intelligence in ways television cannot.

MIT Technology Review (ISSN 1099-274X), July/August 2021 issue, Reg. US Patent Office, is published bimonthly by MIT Technology Review, 1 Main St. Suite 13, Cambridge, MA 02142-1517. Entire contents ©2021. The editors seek
diverse views, and authors’ opinions do not represent the official policies of their institutions or those of MIT. Periodicals postage paid at Boston, MA, and additional mailing offices. Postmaster: Send address changes to MIT Tech-
nology Review, Subscriber Services, MIT Technology Review, PO Box 1518, Lincolnshire, IL. 60069, or via the internet at www.technologyreview.com/customerservice. Basic subscription rates: $80 per year within the United States;
in all other countries, US$100. Publication Mail Agreement Number 40621028. Send undeliverable Canadian copies to PO Box 1051, Fort Erie, ON L2A 6C7. Printed in USA. Audited by the Alliance for Audited Media.
Let
the story
come to
you.
Tech news from our journalists delivered straight to your inbox.

WEEKDAYS FRIDAYS (SUBSCRIBER EXCLUSIVE)

The Download The Algorithm

Navigate the world of tech news Artificial intelligence, demystified

WEEKLY SATURDAYS

Coronavirus Tech Report Weekend Reads

How covid-19 is changing our world Technology in perspective

WEDNESDAYS MONTHLY

The Airlock fwd: Economy

Your gateway to the future Your guide to growth and prosperity
of space technology in the age of technology

technologyreview.com/inbox
 Stopping spyware from slithering into
our nation’s critical infrastructure.

battelle.org/spyware