Professional Documents
Culture Documents
[MUSIC PLAYING]
NARRATOR: This is
Louis XIV, also known
Culture?
Astronomy?
So in the spring
of that year, Louis
oceans.
In the grandest of
human traditions,
Travelers trying
to figure out how
in hundreds of languages,
expecting someone
CREW: Yeah.
[MUSIC PLAYING]
He knows a few
things about search.
moved back to
India to Bengaluru.
my mom.
In 5th grade, I
got two presents--
[MUSIC PLAYING]
it was so dramatically
different from today.
I think that
reduction in friction
is absolutely dramatic,
because it can enable people
information.
So that combination of a
deep technical problem and I
[MUSIC PLAYING]
I was in an office
with Urs Holzle.
Because without
really good ranking,
wasted if nobody
could actually find
CREW: Pandu?
OK.
[MUSIC PLAYING]
PANDU NAYAK: No
query left behind.
NARRATOR: Before
working at Google,
high-level control to a
spacecraft called Deep Space 1,
in my life--
in my professional
life, I guess.
So ranking is important
because if we simply
[MUSIC PLAYING]
at Stanford.
[CHEERING]
SPEAKER 2: And
today, our mission
[MUSIC PLAYING]
CATHY EDWARDS: I
think if we weigh up
Our job is to
filter through that
So information means
really anything.
is important, because
universal means for everyone.
DAVID BESBRIS: We
might be goofy people
who come to work in T-shirts
and desperately need
SPEAKER 3: 1.0.
NARRATOR: So it sounds
like the mission is pretty
[MUSIC PLAYING]
NARRATOR: This is
server rack 3349b.
It lives here in
Ballybane, Ireland,
Repair Center.
Search is a big
piece of software
longer.
Lasagna appears on 59
million of those pages.
Most people
searching for lasagna
Zurich, London,
India, Japan, so on.
ELIZABETH TUCKER:
Anything we need to know?
over a decade.
SUNDEEP TIRUMALAREDDY:
Things are getting exposed.
SPEAKER 4: Part of
that is building--
[DOOR SLAMS]
[MUSIC PLAYING]
a set of mathematical
instructions
to return high-quality
information based on a user's
This is the
information that Google
tries to organize
and make universally
VOICE: We will be
with you shortly.
VOICE: Hey!
NARRATOR: Spam.
CATHY EDWARDS:
Yeah, so let me just
CATHY EDWARDS:
Broadly, spam is what
we consider a low-quality
page that is artificially
NARRATOR: She's
talking about pages
topicalarticles.info,
the kind of websites
Sometimes it's
commercial interests.
CATHY EDWARDS:
Spam, where they're
Or sometimes it can
just be to capture
It makes it even
harder for us and it
recipe-for-Coke-level
guarded secret.
Fighting spam is a
cat and mouse game.
CATHY EDWARDS: As
an example, 40%
[MUSIC PLAYING]
But as an engineer,
you ask yourself,
In fact--
We're coming up
with new devices.
to start with.
roughly a couple of
thousand launches in a year,
expertise, authoritativeness,
and trustworthiness.
so the thousands of
independent evaluators keeping
quality, freshness,
relevance, but we also
So we have to be
careful that we are not
SUNDEEP TIRUMALAREDDY:
In some ways,
INTERVIEWER: I did a
search a couple days ago,
a complicated thing,
three-hundredths of a second.
I mean, it seems
inconceivable you
To people sometimes,
the internet
so to speak.
We're--
PHILLIP: Actually,
this is quite big.
PETRA: [INAUDIBLE]
searches a day
so to speak?
of the internet.
millions of servers
and hard drives
to be able to support
the world's websites.
URS HOLZLE: So each
of these data centers
in South Africa,
you're not sending
And so that's
actually the reason
KEVIN: Yeah.
no paywalls, no
subscriptions, no "you've
BART: Hi.
ALL: Hi!
NARRATOR: Bart
buys ads on Google
BART: Pittsfield!
NARRATOR: --searches,
for instance,
It helps people
find mowers to buy,
been collecting
Google's dumbest Search
[CHEERING]
Senior Software
Engineer Eric Lehman.
So this caused an
overflow error,
Mm-hmm.
Blue?
Sure.
[LAUGHS] Yeah.
Avogadro's number is a
sort of important constant
in chemistry.
enough results to
fill 27 libraries,
Modern History.
JOHN-PAUL GHOBRIAL:
It used to be,
If there's an error
in the printed word,
And so if that
content is out there,
I would consider to be
reprehensible or really
offensive results.
[MUSIC PLAYING]
a case of misinformation,
because the Holocaust did
actually occur.
of misinformation.
We want to make
their lives better
They're talking
about the Holocaust
They tend to be
lower quality even
were overpowering
the quality signals
for users.
PANDU NAYAK: We
have long recognized
incredibly important.
NARRATOR: Misinformation is
one of the challenges that
by guessing what
they're searching
Autocomplete was
designed to help people
First, we developed
a set of policies
sexually explicit
content, hate speech.
we try not to
surface predictions
PANDU NAYAK: We
use those reports
is that this in no
way prevents users
They're absolutely
free to do that.
NARRATOR: Think
about it this way.
With Autocomplete,
it's the kind of door
We do make mistakes.
We want Search to be
very easy for people.
But behind that is an extremely
hard technical problem
in a query, not
just matching words,
things together.
It's an absolutely
fascinating problem
as human beings.
NARRATOR: As long as
there have been machines,
So humans had to
come up with new ways
solved equations.
NARRATOR 2: Holes
punched in the card
represent data to be
placed in the computer.
BEN GOMES: As we
understand language better,
[CHEERING]
we want to be able
to understand what
have available.
[MUSIC PLAYING]
I mean, we began to
work on this problem
Correction.
[APPLAUSE]
at Google.
Today, he is
revolutionizing how we live.
in machine learning.
INTERVIEWER: What
kind of impact do
GEOFFREY HINTON: I
hope that it allows
[CHEERING]
in natural language
processing was announced.
Transformers--
NARRATOR: So BERT's a
big deal for Search.
[MUSIC PLAYING]
ELIZABETH TUCKER: So
I think we're finally
SUNDEEP TIRUMALAREDDY:
That's why
So people use
language every day.
So it's actually
sort of interesting.
Early on in information
retrieval, which
is the science
behind Search, people
So like a lot of
little connector words,
queries in a more
natural way for humans
[MUSIC PLAYING]
ELIZABETH TUCKER: I
would have guessed that--
Months go by.
Progress is slow.
a distinction in
that rank, so I'm
of numbers.
of computing power.
PANDU NAYAK: If
we don't get that,
ELIZABETH TUCKER: I
would say, in general,
So we have these
high-level measurements
experimental results.
I would like us to
go through some wins.
ERIC LEHMAN: It
is a tough query.
Holy cow.
Hi, guys.
I don't know.
[LAUGHS]
So Launch Committee is
essentially the final review
you're essentially
saying, hey, we
into production.
[MUSIC PLAYING]
It may be over, or
it may be approved
Regardless, it's
a big milestone.
Because despite
their best intention,
[INTERPOSING VOICES]
DAVID BESBRIS:
Generally speaking,
Because it can be
a little tough.
think is particularly
worth looking at
PANDU NAYAK: So
one of the things
is to weigh these
wins and losses.
going on in here.
However--
PANDU NAYAK:
DeepRank illustrates
You see--
is a really hard
concept to get in IR.
And so this is
where it can excel.
DAVID BESBRIS:
When it's all done,
the coordinator of
the Launch Meeting
[LAUGHTER]
The decision is to
launch DeepRank.
[SIGHS]
SPEAKER: Awesome.
a little awkwardly
for a few minutes,
And probably
that'll happen here.
It was a pretty
remarkable project.
ELIZABETH TUCKER:
Congratulations.
It feels significant.
ELIZABETH TUCKER: We
will work on that.
[LAUGHTER]
[MUSIC PLAYING]
PANDU NAYAK: Solving
the Search problem
to information than at
any other time in history.
[NON-ENGLISH SPEECH]
accessible to people.
And I think the thirst
will still be there,
looking for.
It enables them
to make decisions
[LAUGHTER]