Professional Documents
Culture Documents
Technologies
P R O F. S H R I K U M A R M
L E A N O P E R AT I O N S A N D S Y S T E M S
SCHOOL OF BUSINESS AND MANAGEMENT
CHRIST (DEEMED TO BE) UNIVERSITY
Mind Map of Chapter Topics
2
The Turing Test
In an influential paper in 1950, Alan Turing, a British mathematician who had proposed
the fundamental basis of computing devices, wrote an answer to the fundamental
question ‘Can Machines Think?’
The Turing test compares the performance of a computer program with that of human
in a live interaction situation.
3
Machine Learning
Backpropagation works on the basic principle of comparing the desired output with the
actual output for some given input data, and then adjusting the weights of the network
by small amounts.
Neural Networks
Learning in Neural Networks
Deep Neural Networks
Induction Learning
Bayesian Learning
Learning from Analogies
Genetic Algorithms
Reinforcement Learning
Limitations of AI and Machine Learning
4
Machine Learning – Neural Networks
d1 w1
d2 w2 Sum of Transfer
w1d1+w2d2+ Function
w3d3
d3 w3
Learning in Neural
Networks
Neural networks are trained with large amounts
of data
Training – Adjustments of weights in the
network
Changes in weights eventually leads to the
desired output
Backpropagation is the way of adjusting weights
starting from the hidden layers connected to the
output layer. Then adjusting weights at the next
hidden layer working backwards connected to
the input layer.
Deep Neural Networks or Deep Learning
Large networks trained with large data sets
DNNs have dozens or more hidden layers
Works on backpropagation
E.g.: Identifying whether an image is of a dog.
Image will be split into many parts – eyes, ears, nose etc.
Each of these images will try to match with data used for training.
Induction Learning
Induction Learning
◦ Induction - Drawing conclusions from data without any prior
hypotheses
◦ Decision trees for classification of data
◦ Decision trees are created based on rules using past data
◦ E.g.: Past data shows the multiple criteria leading to grant of a
bank loan – If the Income is High and Repayment is Good; then
Loan Grant is a Yes
◦ If the Income is Low and Avg Monthly Savings is Low; then the
Loan Grant is a No.
Bayesian Learning
Bayesian Learning
◦ Classification based on probabilities of data belonging to certain classes
◦ Based on Conditional Probability – Probability of Event given a Condition or Probability of Effect given a
Cause
◦ Reverse Probability - Probability of Cause given an Effect
◦ E.g.: Whether an email received in inbox is spam or not. Data consists of email that users classify as
spam or useful.
◦ Software scans the words and phrases in the email and updates the probability of P(email is spam|
indicative words).
◦ When new email is received and scanned, if probability is above a threshold level, the email is
classified as spam
Reinforcement Learning
Idea is that agent has to perform a series of actions that will lead to a goal
Series of actions has to be learnt through rewards and punishments (Reward is 1, Punishment is
0)
Agent knows set of actions and knows the current state, Outcome of actions is not known in
advance and sequence of actions is not known.
Agent observes state of the world, taken an action, the state changes and reward is given to an
agent.
Agent learns which actions to be taken to maximize an award.
E.g. A robot has to place different types of glasses and glasses at specific places on a table and in
a certain sequence in a restaurant.
Learning from Analogies & Genetic
Algorithms
Learning from Analogies
◦ Instance based learning
◦ Finding a match from previous examples to find classification for data
◦ E.g.: For identifying a variety of watermelon, people will look at size, shape, color etc.
◦ Categorising based on some parameters
◦ Method is called k-Nearest Neighbour (kNN) algorithm
Genetic Algorithms
◦ Algorithm follows natural evolution – Reproduction, variation and adaption
◦ Problem representation with the help of alphabetical or numeric strings that are treated as genetic information
◦ Generate offspring strings that have within them solutions to problem
◦ Crossover and mutation done to test fitness to solve a problem
◦ Offspring strings solve problems of classification, prediction etc.
◦ Challenge is to find appropriate representation of a problem
◦ Application – Managing robots, solve logistics problems, designing neural networks
Limitations of AI and Machine Learning
Many AI systems cannot explain how and why they came up with conclusions and answers
Learning systems reflect biases in the data that many distort their performance
AI systems operate in limited domains like object classification, disease classification, market
predictions etc. But they lack general intelligence – like solving a crossword puzzle
Systems of perception such as image recognition and speech understanding work fine in certain
domains but performance degrades in live situations with varying and adverse environmental
conditions.
Implications for Managers
Managers will have to decide on the appropriate division of work between people and machines.
Some tasks, requiring judgement and decision making, will invariably require people; however many
other tasks can be transferred either completely or partly to intelligent machines.
The right balance of this division of work will be determined by each context and circumstance.
Persuading organisational personnel and staff about the challenges and opportunities of AI/ML
technologies will remain as one of the most difficult tasks of managers.
Managers will also have to actively seek new areas of innovation in products and services with such
technologies.
13
Blockchain technology
How does Blockchain work?
https://www.youtube.com/watch?v=SSo_EIwHSd4
https://www.youtube.com/watch?v=3xGLc-zz9cA
Secure, peer to peer network without need for intermediary (e.g. a bank or lawyer or notary)
New additions to transactions or updations are validated by peers in the network before being added to the
chain
Some nodes in the blockchain can do the record validations – earn a financial incentive because of
investment in computing power and time.
Changes that are not valid cannot enter the chain and interested parties get an alert.
Applications of Blockchain technology
Risk Reduction: Home owners currently store customer identities and financial data on their own
servers, which can be hacked and leaked, exposing owners to litigation and large liabilities. On the
blockchain, you needn’t trust a vendor with your data; there is no central database to hack and leak.
There are only individual peer-to-peer pseudonymous transactions.
Case - bAIRBNB VERSUS AIRBNB
Insurance: Today Airbnb offers $1 million insurance for owners and compensates them for theft and
damage. On bAirbnb, owners can get the bAirbnb insurance DApp. Renters with good reputations like
you have lower insurance rates and needn’t subsidize renters who lack caution, scrutiny of prospects,
or poor treatment of property. When you submit a booking request, bAirbnb sends your public key
(persona) to the insurance contract for a quote. The insurance DApp contacts a list of trusted
providers.
Payment Settlement: On the blockchain, you transfer funds to the owner in seconds, not days as
with Airbnb. Owners can manage security deposits more easily with smart contracts. Some parties
use escrow accounts to release payments partially (nightly, weekly, hourly, etc.) or in full as the
parties agree. In disputes involving smart contracts, parties can call for arbitration.
Case - bAIRBNB VERSUS AIRBNB
Property Access Using Smart Locks (IoT device): A smart lock connected to the blockchain knows
when you have paid. When you arrive, your near-field communication-enabled smart phone can sign
a message with your public key as proof of payment, and the smart lock will open for you. Owners
need not drop keys off to you or visit the property unless they want to say hello or address some
emergency.
Blockchain technology – Global Supply
Chain
◦ Business Challenges
◦ Trusting all players and partners in many transactions
◦ Supplier ensures that goods being imported are actually shipped
◦ Relevant papers have been included – Documents related to local customs and local shipping
regulations
◦ Payments have been received
◦ No information has been leaked
◦ No fraudulent entries have been made
◦ Global supplier knows exactly when a shipment is made as this is recorded on the blockchain
◦ Smart contract enables timely payments – funds released as soon as entry is made
◦ When shipment is made, relevant documents are included in the ledger entry
◦ All correct documents received by the correct parties at the right time
Blockchain technology – Supply Chain
◦ Complete transparency of transactions
◦ Certain goods which need to maintained, under temp and pressure specifications, are also recorded on
the blockchain (either manual entry or IOT)
◦ Upon receipt of goods, buyers can verify receipt as well as conditions of transport.
29
Internet-of-Things
Industrial IoT or IIoT or Internet 4.0 Manufacturing and service operations are monitored and controlled.
Harvested data is analysed to identify improvements. Even 1% gain in efficient energy usage results in huge
cost savings.
Rolls Royce has changed business model of Jet Engines
Company used to sell Jet engines to airplane manufacturers and then setup maintenance contracts to service
those engines
With IoT, this model has changed: RR embeds thousands of sensors in its engines to monitor every aspect of
the engine and send data to the company over the internet.
Data is used with complex software to simulate performance of engine using a ‘digital twin’. Company knows
how much engine has run, wear and tear of parts, when parts have to be serviced etc.
Customers pay for service of engine – usage and not for the total engine itself.
Shift from cost of buying assts upfront to paying for engine service on a measured basis.
30
Blockchain enabled IoT – Ledger of
Everything
Nine network features:
Resilient: Self-corrects; no single point of failure
Robust: Can handle billions of data points and transactions
Real-time: Stays on 24/7/365 and data flows instantly
Responsive: Reacts to changing conditions
Radically open: Constantly evolves and changes with new input
Renewable: Can be multipurpose, reused, and recycled
Reductive: Minimizes costs and friction, maximizes process efficiency
Revenue-generating: Enables new business models and opportunities
Reliable: Ensures integrity of data, trustworthiness of participants
Animating Things – 12 Disruptions
Transportation
In the future, you’ll call up an autonomous vehicle to get you safely where you need to go.
It will intuitively take the fastest route, avoid construction, handle tolls, and park all on its own.
In times of traffic congestion, your vehicle will negotiate a passing rate so that you arrive at your
destination on time, and
freight managers will use the blockchain-enabled IoT on all cargo to clear customs or other
required inspections quickly.
Animating Things – 12 Disruptions
Infrastructure Management
Many professionals will use smart devices to monitor location, integrity, age, quality
Any other relevant factors of pavement, rail lines, power poles and lines, pipelines, runways,
ports, and other public and private infrastructure
To monitor conditions, detect problems (e.g., breakage or tampering), and initiate a response
both rapidly and cost-effectively.
Animating Things – 12 Disruptions
Energy, Waste, and Water Management
“Send a truck to empty me,” said the overflowing waste bin.
“Fix me,” said the leaky pipe.
The Internet of Things should inspire a hundred new children’s books.
Traditional utilities in both the developed and developing world can use the blockchain-enabled
IoT for tracking production, distribution, consumption, and collection.
New entrants without significant embedded infrastructure are planning to use these
technologies to create entirely new markets and new models (e.g., community microgrid)
Animating Things – 12 Disruptions
Resource Extraction and Farming
This technology can also help track expensive and highly specialized equipment and make it more
widely available for just-in-time usage and cost recovery;
Improve miner and farmworker safety through tagging of safety equipment and automated
checklists (to ensure that equipment is being used properly);
Monitor weather, soil, and crop conditions to start irrigation, automated harvesting, or other
actions; and
Compile “infinite data” analytics to identify new resources or advise on agricultural best practices
based on past patterns and results.
Sensors in soil and on trees could help environmental protection agencies to monitor farmers and
their usage of the land.
Animating Things – 12 Disruptions
Environmental Monitoring and Emergency Services
Monitoring air and water quality and issuing alerts to reduce pollutants or stay indoors;
Flagging dangerous chemicals or radioactivity for emergency workers;
Monitoring lightning strikes and forest fires;
Installing earthquake and tsunami early warning and alert systems; and storm monitoring and
early warning.
In addition to improving the response time for emergency services and reducing the risk of
these events to human life, we could use this data to increase our understanding of underlying
trends and patterns, identify preventive measures in some cases, and improve our predictive
capability to provide even earlier warning.
Animating Things – 12 Disruptions
Health Care
In the health care sector, professionals use digitization to manage assets and medical records,
keep inventory, and handle ordering and payments for all equipment and pharmaceuticals.
Today, hospitals are full of smart devices that oversee these services, but few communicate with
one another or take into account the importance of privacy protection and security in direct
patient care.
Blockchain-enabled IoT can use emerging applications to link these services.
Applications in development include monitoring and disease management (e.g., smart pills,
wearable devices to track vital signs and provide feedback) and improved quality control.
Animating Things – 12 Disruptions
Financial Services and Insurance
Financial institutions could use smart devices and the IoT to tag their claims on physical assets,
making them trackable and traceable.
Because digital currencies enable the storage and transfer of value rapidly and securely for all
users large and small, they also enable risk assessment and management.
Animating Things – 12 Disruptions
Document and Other Record Keeping
Physical assets can become digital assets.
All documentation relating to a particular “thing” can be digitized and carried on the blockchain
including patents, ownership, warranties, inspection certification, insurance, replacement dates,
approvals, et cetera, significantly increasing data availability and integrity, reducing paperwork
handling, storage, and loss, and other process improvements related to that documentation.
For example, a vehicle will not start if its owner’s driving license is not renewed or suspended, if
its liability insurance has expired, if its owner has failed to pay parking fines or moving violations.
Items on the shelves will notify store managers when they’ve passed their “sell by” date. Store
managers might even program these items to lower their own price as the sell-by date
approaches.
Animating Things – 12 Disruptions
Industrial Operations—The Factory of Things
The global plant floor needs a global Ledger of Things, aka the industrial blockchain.
Factory managers will use smart devices to monitor production lines, warehouse inventory,
distribution, quality, and other inspections.
Entire industries may adopt the ledger approach to significantly increase efficiency for such
processes as supply chain management.
Large and complex machines, like airplanes and locomotives, consist of millions of parts.
Each individual component of a jet engine or railcar could have sensors that send out an alert
when it needs fixing.
Animating Things – 12 Disruptions
Retail Operations and Sales
Walking down the street, your mobile device advises you that the dress you love is available at
the store.
Walk into the store and the dress, in your size, is waiting for you.
After trying it on, you scan it and the payment is complete.
But you’ve got other things to do, so the dress is delivered to your house before you get home.
In addition to operational efficiencies and environmental monitoring, retailers will be able to
personalize products and services to identifiable customers as they walk or drive by based on
their location, demographics, known interests, and purchasing history, provided that those
customers opened their black boxes to retailers on the blockchain
Seven Conditions for Success
Incubators and Entrepreneurship: Innovation is nurtured in environments established exclusively for
this purpose.
Corporate Leadership: Centers of blockchain innovation very often enjoy close ties with established
business communities. In areas where corporate entities manifest a culture of curiosity and market
positioning as innovators, blockchain developments can especially thrive.
Educational Institutions: Great computer science schools can lead to great blockchain innovation
ecosystems.
Investment Climate: Angel capital, venture capital, and strong financial services industries must have a
risk tolerance for this kind of innovation.
Government Support: One of the most important things that governments can do is to be model users
of the technology itself. Government community-based initiatives can also fund innovation directly. In
Hangzhou and Guangzhou, China’s government is pouring billions into blockchain development
Seven Conditions for Success
Government Support: One of the most important things that governments can do is to be
model users of the technology itself. Government community-based initiatives can also fund
innovation directly. In Hangzhou and Guangzhou, China’s government is pouring billions into
blockchain development.
Regulatory Environment: Governments can create balanced policies that neither over-regulate
or under-regulate.
Communities of Talent: Highly educated populations are an important factor for innovation to
take root in any given jurisdiction. How do you initiate a national brain gain rather than brain
drain?
Mind Map of Chapter Topics
44
Cloud Computing
45
Cloud Computing
46
Cloud Computing
Cloud services pricing can be based on usage and needs of the client.
47
Cloud Computing
Elasticity
Ability of the IT infrastructure to expand and contract as per client demand.
Hosted facilities – Components that are acquired become assets and are permanently available for use.
When demand declines, utilization of these assets decline
Measured Use
Hosted services assemble components from different brands. Difficult to monitor usage in a comprehensive manner
Cloud provider can give detailed reports of usage like CPU utilization, memory usage, data usage for read or write etc.
48
Cloud Computing
49
Cloud Computing
50
Cloud Computing
51
Virtualization
Use of virtualization allows many CPUs to be function as a single unit on which many Operating systems
and applications can be run simultaneously. This is used to increase capacity and improve services.
52
Cloud Computing
Types of Cloud Infrastructure
Private Cloud: A private cloud is meant for a single client whereas a public cloud is
shared by many clients.
Private cloud consists of servers that are rendered as a cloud through virtualization. Built of
commodity hardware at low cost. Higher security since infra is within the organization.
Disadvantage: Manage the cloud on your own.
Private cloud can also be done by a cloud provider.
53
Cloud Computing
Cloud Service Models
Infrastructure-as-a-Service (IaaS): IaaS provides data centre facilities
Platform-as-a-Service (PaaS) : PaaS provides a software development facility
Software-as-a-Service (SaaS): SaaS provides applications over the internet.
54
Cloud Computing
Cloud Service Models
Feature IaaS PaaS SaaS
Typical Usage Infrastr for multiple Build and deploy new Single application for
servers and applications applications; maybe for focused use
new markets
55
Cloud Computing
Challenges of Migrating to the Cloud
◦ Tight coupling between hardware and software : Sw works on specific hardware only.
◦ The cost is not justified : Legacy systems may be cheaper due to depreciated usage.
◦ Security is a key concern : Some buyers are worried about hacking and other internet based
attacks.
56
Cloud Computing
5 Major Reasons for Adopting the Cloud
◦ Applications that have volatile demand: Potentially sudden increase or decrease in demand
◦ Low business risk and commodity applications: Save management time and cost
◦ Applications that require collaboration: Sharing of data and files across regions.
◦ Capital available for infrastructure is low: Start-ups, large scale and high performance can be acquired
57
Cloud Technologies
Software Defined Networking (SDN)
SDN: Virtualising networking tasks
Problem: Routers and switches in a hierarchical configurations
Problem: Networking devices need to be reassigned to other CPUs
Control plane: Best routing of data packets over the network
Data plane: Physical routing of data
Control plane is free to solve routing problems over the network
58
Cloud Computing
59
Cloud Computing
60
Internet-of-Things
IoT are sensors and actuators that are connected via the internet, for responding to and
control of events.
61
Internet-of-Things
Industrial IoT or IIoT or Internet 4.0. Manufacturing and service operations are monitored and controlled.
Harvested data is analysed to identify improvements. Even 1% gain in efficient energy usage results in huge
cost savings.
Rolls Royce has changed business model of Jet Engines
Company used to sell Jet engines to airplane manufacturers and then setup maintenance contracts to service
those engines
With IoT, this model has changed: RR embeds thousands of sensors in its engines to monitor every aspect of
the engine and send data to the company over the internet.
Data is used with complex software to simulate performance of engine using a ‘digital twin’. Company knows
how much engine has run, wear and tear of parts, when parts have to be serviced etc.
Customers pay for service of engine – usage and not for the total engine itself.
Shift from cost of buying assts upfront to paying for engine service on a measured basis.
62
INTERNET-OF-THINGS
Examples of IoT
63
INTERNET-OF-THINGS
64
INTERNET-OF-THINGS
Sensors
A sensor is a device that detects events or changes is its physical environment and provides an
electronic output.
E.g.: Temp, pressure, flow, level, Image, Noise, Air particle etc.
Choice of sensor: Data filter, power consumption, size, sensitivity and accuracy.
Actuators
Actuators are tools or mechanisms that can change their own state or the state of other devices
and machines.
Types – Electrical (electrical signals to mechanical motion), Mechanical (Movement of gears or
levers), Hydraulic and Pneumatic (Devices that respond to pressure in liquids and gases)
65
INTERNET-OF-THINGS
IoT Architecture
• Security
• Privacy
• Internet access
67