WORKSHEET 2: Al or not No (if t is not Al; or May

be (where it can
is Al;
and which are not? Choose Yes (ifit
Which of the following are Al
be both, depending on the viewpoint).
Microsoft Access
Database for Yes No May
a be
A web application that
accesses

Q1.
to the queries asked by the users.

information pertaining
finding
in the past May
system that looks for patterns Yes No
Q2. A stock price forecasting in the future.
e

data for predicting the stock prices

s e n s o r in the
classroom

Q3. A s m a r t attendance system

uses a fingerprint
Yes No May
to date records to parents, be
for recording attendance, and provides up
on request.
teachers, and school authorities
user data to provide May
Q4. A smart recommendation system that analyses Yes No
recommendations to its customers.
Its recommendations be
the purchase
are based on customer's past purchases.

Q5. A smart online video storage system, which

stores user videos,
Yes No May
and categorises them based on some be
generates automatic tags,
standard parameters.

Q6. A smart security system that can identify the

authorised persons based
Yes
May
on their face, voice, behaviour, etc.
No be

sOLUTION:
1. No 2. Maybe 3. No 4. Yes 5. Yes 6. Yes

THREE DOMAINS OF ARTIFICIAL INTELLIGENCE

Artificial intelligence is generally divided into three separate domains:
Data
Computer Vision (CV)
Natural Language Processing (NLP) Data

DATA
Natural
Every Artficial Intelligence System- irrespective ofits function, nature, Langua
or capability- is heavily dependent on data for its functioning or in other ooeasi
words data is at the core of every Al system. Both of the other Al domains
also need data for their functioning Data is also at the core of the General
Al systems, as these systems will have the capability of processing data for
learning and growing8
Artificial Intelligence
Data for artificial intelligent system means data that is relevant for the Al system in question. For example
an Al-based automatic student attendance system for school will not be helped by the data of images

Kips I ARTIFICIAL INTELLIGENCE

134
of world leade.ders. Different type of data serves different purposes in artificial intelligent systems. We
will
nt

more
about lata in Chapter 2.
study

E RELATION BETWEEN DATATYPE AND ITS USAGE IN AI SYSTEMS

THE RELATION

Audio Data Video orimage Data Textual Data Big Data

Voice Recognition Computer Vision Classification systems Predictions
Natural Language Natural language Insights
Learning learning Forecasts
Decision Making Decision Making

cOMPUTER VISION (CV)

This domain of Artificial Intelligence is working towards the development of Al systems, which will be
able to perceive the human worlds as human beings do. There has been sustainable development in this
domain, and this technology is currently being used in a number of Al-based systems.

USES OF cOMPUTER VISION

Face Recognition ldentifying faces in images and videos

Applications like Google Photos, Snapchat, etc.
Social media networks like Facebook, Instagram, etc.
Law enforcement agencies like Interpol, FBI, etc.

CBIR Content-based Image Retrieval: Identifying images based on their

composition, colour, texture, etc.

Search engines like Google and Bing

Medical image databases of CT, MRI, etc.
Scientific databases like Earth Sciences
Smart Interactions One more way of providing inputs to the computer systems

Gaming systems like Microsoft Kinect

Games like Emoji Scavenger Hunt
ystems for differently-abled individuals
Environment Perception Analysing videos, images or video feeds for identifying patterns and
perceiving environment

behaviour
aw enforcement agencies for identifying illegal/suspicious
Home Security Systems
Office Security Systems
Drone-based Surveillance Systems
Smart vehicles

NATURAL LANGUAGE PROCESSING (NLP)

domain of Al is working towards the creation of artificial intelligence systems, which will be capable
of nunicating with human beings using natural language rather than by syntax
or identification of
keywords. This dor
iomain is working for the development of both oral and spoken languages.

ARTIFICIAL INTELLIGENCE I Kips 135

 cOMPONENTS OF NLP

Natural Language Processing contains two main components:

written language, which
Natural Language Understanding (NLU):
For understanding spoken or
1.
includes:

establishing linkage wich natural language inputs and what they represent.
analysing different aspects of the language.
and in the
2. Natural Generation (NLG): For
Language producing meaningful phrases sentences

form of natural language, which involves:

Text planning: Retrieving relevant text from the data stores

Sentence planning: Deciding on the correct words, linking them into meaningful phrases, etc.

Text realisation: Combining phases and words for forming sentences

For Al systems, NLU is much more complicated than NLG. Interestingly, human beings find NLU easier
than NLG.

WORKSHEET 3: Identifyingthe Domain of Al

Identify some (ot least 5) of the unique software and solutions which are reloted to the domain of artificia
intelligence. Feel free to use the internet, and ask your parents and teachers. But do not ask your fellow students.
The first row has been filled as an example.

S. No. Data Computer Vision Natural Language Processing

Example Google Search Google Image Search Google Now
1.
2.
3.

7.
3.
.
10.
11.
12.
13.
14.
15.

136 KiPs | ARTIFICIAL INTELLIGENCE

RELATE

THERELEVANCE OFAIIN DAILY LIFE

SIX DIMENSIONS OF AI
The relevance of Al in the day to day human life can be seen through the six dimensions of the artificial
intelligence.

AUTOMATION Automatic repetitive learning and discovery by Using data

AUGMENTATION Augmentation of the intelligence ofexisting products

ADAPTATION Adapting by using data

ANALYSIS Finding deepertruths by analysing data

ACCURACY Increasing accuracy of existing or new systems

ACQUISITION Acquiring more out of data

AUTOMATION
In the pre-Al era, automation was limited to hardware-based robotic automation. For example, a stamping
machine was automatically stamping envelopes. The advent of the Al changed this and ushered in the
automation that uses data for learning and discoveries. This created solutions and products, which
improved over a period of time. For example, Google Pixel smartphones use Al technology for automating
its camera. This means that the camera automatically adjusts different parameters and in some instances,
takes multiple shots for delivering the best consumer experience. This automation continuously learns
for the data gathered by it and keeps on improving itself. This means that the quality of photos taken by
Google Pixel users will continue to improve with the usage.

AUGMENTATION
The development of new Al-based products is rare. Organisations across the globe instead prefer to use
Al technology for augmenting their existing services, solutions, and products. For example, Google Pixel.
Google did not develop any new product. Google simply used the existing product- Mobile Phone with
Camera- and used Al technology for augmenting the capability of this product.

ADAPTATIiON
Al systems do not execute programmed instruetions. Infact, these systems adapt or program themselves
by learning from the data fed to them. This leads to the creation of products and services which keep on
adapting themselves to serve the users better. For example, smart thermostat controls keep on learning
rom the user behaviour and environmental data for providing the best possible experience.

ANALYSIS
Some Al data models like neural networks use multiple layers for processing data (discussed in detail
in Chapter 3). The insights gained from the data becomes more in-depth with the addition of more

ARTIFICIAL INTELLIGENCE | KiPs 137

 For
example, fraud
layers. This, in turn, increases the understanding and leads to better predictions.
that had escaped detection earlier.
detection systems keep on analysing data for uncovering hidden frauds

ACCURACY
The accuracy of Al systems is more for tasks that require insights, decision making, or analyse a large
amount of data. For example, the results provided by Bing search engine continuously improve inaccuracy.

ACQUISITION
Organisations across the world have collected massive amounts of data. This data already has solutions
formany problems and situations. We have just not acquired these solucions. The Al systems can look for
solutions in this data. For example, we are unl ocking the secrets contained in Human Genome.

SMART LIVING: SMART HOMES TO SMART CITIES

Artificial Intelligence is changing the way in which we experience
life by changing our home, work, and living environment. DEFINITION OF IMPORTANT TERMS
Buildings are becoming smarter, i.e., using Al technology for Smart Home: Houses that use smart
improving the lives of their inhabitants. These buildings are devices.
Smart Buildings: Any building (includ-
connecting together to create smart cities. Artificial intelligence
ing homes) that uses smart devices.
systems use the data collected from the various smart devices
Smart Devices: An electronic device
for learning and improving both services and products.
connected to and is capable of commu-
These smart cities are indispensable for the sustainable devel- nicating through the internet. These
opment of the Earth when the pace of urbanisation is increasing devices can also communicate among
across the world, and the
availability of resources is themselves.
decreasing. Smart Cities: Cities that (i) have smart
ADVANTAGES OF SMART BUILDINGS buildings and (i) use smart technology
The various functions of the smart buildings for improving the lives of the citizens.
can be regulated
from anywhere across the world.

They provide a safe environment for children and adults

by using various surveillance and
measures.
securit
They can control the environments
of the home for providing inhabitants with the best
environment. possible

ADVANTAGES OF SMART CITIES

They can better serve the citizens residing in them by providing better services.
They can use the data collected from smart
buildings for improving the services provided by them.
They can interact witch the smart
buildings for providing end to end citizen services.
They can help in reducing resource usage.
They can help in the sustainable development of the environment.
PHASES IN DEVELOPMENT OF SMART CITIES
PHASE 1
In this phase, the
concept of smart cities was not connected with the of Al
of smart cities was understood to be
limited by providing community
use
technologies. The concept
services to the citizens
using any
138 Kips | ARTIFICIAL INTELLIGENCE
 the
chnology. For
Information Tech. example, replacing paper water bills with digital water bills or providing
bill payments.
for online
facility

PHASE II
hic phase, the use of Al technology started. The advent of smart devices coupled with Artificial
In t
arelligence allowed smart cities to analyse different aspects of the city for improving them. For example
trafft light sensors that change the traffic lights depending on the traffic.

PHASEII
This is the latest phase in the development of the smart cities. This includes increasing citizen involvement

in the
decision making process.

DATA FOR SMART CITIES

We have already seen how data is one of the three domains of Arcificial Inteligence. Smart homes and
buildings are made by interconnected smart devices. These devices gather a lot of data when they are
running. This data helps smart homes, buildings, and cities to become smarter. Let us see how smart cities
can use this data for improving the lives of the citizens:

SMART CITIES AND SMART DATA

Smart devices collect and communicate a substantial amount of data;
many times this data is collected
and transmitted in a real-time.
Following are the three examples that show the smart cities using the
collected data:

Electricity: Various smart devices in the smart

building keep a track of energy
helps them in reducing consumption during different times and situations. consumption
and usage.
This data
This data from individual
buildings can be used by the cities for regulating the power generation and
to substantial supply. These initiatives can lead
savings for cities.
Maintenance: The smart buildings are capable of
products need servicing and replacement. This data self-monitoring,
i.e., identifying when some of their
be used by manufacturers and service
can
tor
reducing their warranty and after sale costs. The citizens of the smart cities, in turn, providers
reduced product failures. benefit from the
Health: Smart devices can
0ata can
gather details of vital signs directly from the users in a non-intrusive way. This
be provided to doctors and
hospitals. The various health care providers can monitor this data for
ensuring better quality of life for the
elderly, disabled, and sick in the city. This can also
ene load on
health infrastructure and the insurance costs. help in reducing
SMART
CITIZENS:CONNECTING INDIVIDUALS WITH COMMUNITY
nart buildings and smart cities create an ecosystem in which the connection
between the individuals,
mmunity, and the state as a whole improves considerably. For example, most of the democracies
their
orld- including India- work on the basis of the in the
he legislative body. These elected representatives representation. The people elect the representativees
make decisions on behalf of the
mart cities provide every citizen of the city a direct say in the decisions persons electing
ed by the cities affecting them. his is alsoo
ensures better co providing updated information to the citizens on the
various aspects of the
onnections between individuals and the communities city. This
which they
Another
er way ir which smart cities create better
way in
live.
encour relationships between individuals and
uraging sustainable development and smart
living communities is by

ARTIFICIAL INTELLIGENCE Kips (139

 with citizens are
in which smart cities connect
Five significant aspects
BUILDING
Smort buildings
UTILITIES utilze monltoring
A smart utlity grid devices that tract
will empower end USage and
TRANSPORTATION
Users fo be more empower users and
Through sensors
aware of their service providers
embedded in
SAFETY energy Users, and to better control
Structural health roadways and and reduce
street lights, real allow utllity
ENVIRONMENT monitoring of componls to electricity
time transit and
With new sensor buildings, bridges dellver only as demands.
networks the and dams as well traffic can be
managed for the much energy or
acCurate monitoring as advanced
purpose of water as required
of environmental waning systems in
conditions like reducing travel
emergency
pollution levels, situations can nowu time and fuel
wildlife counts, andd be put in place. inefficiencles
woter runoff all
become possible.

SMART LIvING - MULTIPLYING THE BENEFITS FOR EVERYBODY

The effect of the smart cicizens linked with the smart communities and cities creating an ecosystem that
leads to the smart living. This smart iving is a complete transformation of the way inwhich we live. Few
of the benefits of these living are:

Regulating environment: Smart cities have sensors that allow the smart buildings to continuously
monitor environment. The data collected from these sensors can
help buildings in automatically dealing
with many environmental conditions. For example, in case of increase in air
pollution levels, the smart
buildings can close windows and doors in order to take air purification measures. These buildings can also
alert emergency services and take reduction measures in of fire, gas
case leak, earthquake, etc.
Emergency services: The city or even region-wide warning systems will be able to provide
warnings and emergency tackling information in case of environmental or other disasters, like emergency
floods, storms, etc. During rescue operations, cities can continue to receive earthquakes,
information from smart
buildings by managing
and prioricising rescue efforts. The various interconnected devices can also start
emergency measures, i.e., automatically re-routing traffic, shutting down risky services to connected
areas, updating individuals with status of the relatives and friends, etc.
Information alerts: Governments at various levels can provide round the
clock information to the
citizens. For example, District collectorate can
update regarding the critical decisions
or traffic
department can provide updates regarding the accidents and traffic jams. affecting the citizens

140 Kips I ARTIFICIAL INTELLIGENCE

PURPOSE
USINGAI TOACHIEVE SDGs
UNDERSTANDING THE SDGS WHAT IS 'SUSTAINABLE DEVELOPMENT?
mean
The term 'Sustainable Development'
can
Sustainable Development Goals (SDGs) are the
in different
set of 17 goals that were adopted at the
United different things to different people
the precise definition of the
Nations Conference on Sustainable Development situations. This makes
used
in Rio de Janeiro in 2012. The SDGs were adopted term difficult. One of the most frequently
'Sustainable
definition of sustainable development is:
to meet the urgent challenges
of the world covering that meets the needs
development is the development
the following three dimensions: the ability of
of the present without compromising
own needs'
1. Environmental future generations to meet their
definition can be broken into the following
This
2. Social components:
satisfies the present needs.
1. The development that
3. Economical allow future generations
2. The development that will
to meet their needs.
the sustainable
The issues covered under
development goals include:

ending extreme poverty.

education.
children receive proper
ensuring all
achieving equal opportunities for all. the planet cleaner
for consumption and production that will help making
promoting better practices
and healthier.

OURWORLD
GOALS TO
TRANSFORM
17

SUSTAINABLE G A L S
DEVELOPMENT

OODHET
QUALITY
TOUCATIO
EER

EAIT 6
AND WELLENG
1 PoY

DECENT WORK AND

UNLATYINVATG
AND NTAETRUGTU reuuTES 12
7 FCONOMIC OROWTH

M PARTNERSHIRS

15 N
1 AUCEJISTICE
6
AND STRON
ISTuuok
17 FOR THE GOAL1S
SUSTAI4ABLE
18 DEVELOPHENT

GOALS

KiPs (143
ARTIFICIAL INTELLIGENCE I
Goal l: End poverty in all its forms
everywhere
Goal 2: End hunger, achieve food
security, and improved nutrition and promote sustainabie agricuture
Goal 3: Ensure healthy lives and promote well-being for all ages
Goal 4: Ensure inclusive and
equitable quality education and promote lifelong learning opportunities
for all
Goal 5: Achieve gender equality and empower all women and girls
Goal 6: Ensure availability and sustainable management of water and sanitation for all
Goal 7: Ensure affordable, reliable, sustainable, and modern energy for all
access to

Goal 8: Promote sustained, inclusive, and sustainable economic

and decent work for all growth.
full and productive employment
Goal 9: Develop quality, reliable, sustainable and resilient infrastructure, including regional and
trans-border infrastructure to
support economic development and human
focus on affordable and equitable access for all well-being with
Goal 10: Reduce
inequality within and among countries
Goal : Make cities and human settlements
safe, resilient, and sustainable
Goal 12: Responsible
consumption and production; achieving economic growth and sustainable
opment requires that we urgently reduce our devet
ecological footprint by changing the way e
produce and consume goods and revenues
Goal 13: Take urgent actions to combat
climate change and its
impacts
Goal 14: Life below water; the world's
oceans-their
global systems that make the Earth habitable temperature, chemistry, currents. and life-driven
for humankind
Goal 15: Life on land:
protect, restore,
and promote sustainable use of terrestrial ecosystems.
manage forests, combat desercification, halt and reverse land
ecologically
loss degradation, and halt biodiversity
Goal 16: Promote peaceful and inclusive societies for
sustainable development. provide access to justice
for all and build effective,
accountable, and inclusive institutions at all levels
Goal 17: Partnerships for the goals; the SDGs can only be realised with a
strong commicment to global
partnership and cooperation

HOW AI IS HELPING TO ACHIEVE SDGs

These SDGs were adopted for
world. Artificial intelligence
meeting the environmental, social, and economic concerns of the
technologies can substancially contribute towards
Development Goals and making the Earth a better place for all of its inhabitants.achieving the Sustainable
The following are some of the
examples of Artificial Intelligence helping
Development Goals. in achieving Sustainable
Ensure availability and sustainable management of water and sanitation for
Sustainable Development Goal 6 deals with the sustainable all
Smart Water Management' solutions management of water and sanitation. Al-based
using smart pipes, soil sensors, smart
rainwater harvesting, smart water meters, smart
water flow control
irrigation management, smart
software, etc. are helping in attalning
144 Kips I ARTIFICIAL INTELLIGENCE
 goal. Thes
These solutions have decreased water consumption by as much as one-sixth of the average
goal.
of this
consumption.

clean energy
Affordable and
Al-based
seventh Sustainable
the seventh
is the
This is Development Goal. In this area, the xperts are expecting
r t energy solutions using smart grids, smart appliances, smart energy storage, smart merers
smart

tne
cors,
ser
smart demand regulations, etc., to cause minimum energy saving of 1.3 billion Mwh by
year 2030.

and communities
Sustainable cities
This is the SDG
Goal-l1. It includes disaster management. One of the best examples of using artificial
intelligence for this goal happened in 2017 during the Hurricane Harvey.
The property was
This hurricane did substantial damage across the Houston area of the United States.
damaged, some roads were flooded, people had to be relocated, etc.
Some roads in Houston were flooded, and some other roads were not. This was severely restricting
disaster management work. Artificial intelligence used satelite imagery data along with the object
detection techniques for creating the real-time map of flooded and non-flooded areas. This allowed
disaster management teams to plan better and coordinate the rescue attempts.

Responsible consumption and production

The artificial inteligence based smart systems like smart farm management, smart irrigation. precision
agriculture, etc, are helping in achieving the Sustainable Development Goal 12 by boosting farm productvity
and reducing wastage.

Climate Actions
In the area of Sustainable Development Goal 13, different Al-based solutions are working to
reduce the
for reducing pollution, for encouraging the
emissions responsible global warming,
of greenhouse gases for
adoption of non-polluting renewable energy sources, etc.
These are few of the Al-based initiatives that are helping in making the development of the Earth sustainable
and the life of its habitants better. Widespread adoption of these
measures give can
TOrts towards achieving sustainable development goals. But the adoption of these technologies is not
substantial
jump to
asy. The following are two of the most important reasons behind this slow adoption:
) Problems with data availability and accessibility: Data is one of the three primary domains
data. The problem is that the
of Al technologies, and Al systems cannot function without adequate
data in the world is not free or easily accessible. To complicate the matter further, the collection and
retention of this data can raise serious privacy and ethical concerns (will be looked at in the next

section).
Problems with the availability of talented manpower: The Al technoloEy is a relatively new
technology. Due to this, there is the shortage of people capable of developing, maintaining, and

mproving Al capabilities, models, solutions, etc.

ARTIFICIAL INTELLIGENCE KiPS 145

POSSIBILITIES

APPLICATIONS OFA INVARIOUS FIELDS

EVERYDAY INFLUENCES OF AI GLOBALLY
we have seen earlier in this chapter, the emergence of the Artificial Intelligence in the middle of the
As
20th century has changed our lives substantially. People of the early 20th century and even of the early and
middle part of the late 20th century will be unable to dream of the life that we live today. We have also
ceen in the last section how the shortage of talented persons in the field is restricting the adoption of the
Al SYstems. In the current section, we will look at the kind of opportunities related to this field.
The field of study required for undertaking Al related jobs depends on the category of the job. But, most
of these jobs will require good knowledge of maths and/or computers. People can enter into entry-level
jobswith a simple graduate degree. The advanced level jobs -particularly the ones related to research and
data modelling require post-graduation at the very least, but doctorates are preferred.

REAL-LIFE EXAMPLES OF THE USE OF AI TECHNOLoGY

Hello Barbie: Hello Barbies are the range of toys that use three concepts of Al:
Language Processing, Machine Learning, and Advanced Analytics for satisfying the
needs of the children. Each of the Hello Barbie is equipped with a microphone. This
microphone records what the child says and the Hello Barbie sends this information
to the Al servers. These servers analyse what has been said, select from the pre-set
dialogs (more than 8000 lines), and transmit the response to the Barbie.
IBM Watson: Watson is a suite ofenterprise-ready Al services, applications,
and tools provided by IBM. lt can be used in multiple ways to develop
different products and/or applications.
Chef Watson is an Al Chef that can develop recipes on its own and help
human chefs create new and unique flavours. Watson Beat provides
BM Watson composers with new inspiration by creating themes and musical elements.It
can also help musicians understand the need for their audience better.

vision: Infervision is a Chinese product that uses Artificial Intelligence

Saving lives. In China, there is a shortage of radiologists. The country
t e s around 1.4 billion CT scans annually for lung cancer alone. The
shortay of radiologists
ofge means that each radiologist has to review hundreds
INFERVISION
cans daily. Infervision elps the radiologists to detect cancers more
accurately and efficiently.
DeepMind: Google created DeepMind with an inspiration from Neuroscience
The idea was to create an Al system that is able to mimic human brains. Google's
DeepMind has successfully defeated human players in a number of games. Google
is expanding this system to healthcare sector. Here, Deep Mind is reducing time
required for planning treatments and helping diagnosing sicknesses.
DeepMtric Volvo: Volvo is a car company that uses Al for replacing

components and servicing

vehicles. Volvo collects data from its
voLWo
is fed to the Al
in a number of ways. This system
ems for helping Volvovehicles
maintain its impressive safety record. The Al system can help

ARTIFICIAL INTELLIGENCE Kips 1417

 the driver and passenger
stressful situations and increase
Volvo to improve the vehicle's performance in
data for researching and developing
self-driving vehicles.
experience. Volvo is also using this
decisions from the vehicle design stage to
BMW: BMW uses data and Al for making
M BMW is also working on self-driving vehicles
the sales and after-sales servicing stage.
cars that do not require any human
lt has set a target of producing self-driving
intervention by 2021.

It uses Al
John Deere: John Deere uses Al technology in the agriculture sector.
on the crop and a need for
systems that use visual data to find out if there pest
is
that use
applying pesticide. John Deere also provides automated farm vehicles
GPS systems and Framsight (Al) for automatically ploughing and sowing the fields.
JOHN DEERE
Talking with Machines: The BBC has used Al
technology to create an audio drama "Talking with Machines." This Al-powered

BBC drama allows listeners to have a two-way conversation with the AI. This allows
the listeners to answer questions and add to something to the story. The audio
drama currently works with smart speakers Amazon Echo and Google Home.

RADAR: UK news agency Press Association (PA) is harnessing the power of artificial intelligence and
robots for delivering local news under project Reporters and Data and Robots (RADAR). This project
involves Al writing 30,000 local stories monthly. These stories are
based on the data from sources like government, public services,
and local authorities. Natural Language Technology is used for
writing these stories
RADAR
Netflix: Netflix is using data and Al technology for predicting what its customers will
enjoy watching. This data helps Netflix in its shift from the content provider to content
creator. The investment in new creations are made on the basis of the Al recommendations.
This also allows Netflix to commit to mulciple seasons of the new shows.
NETFLUX
Walmart: Walmart is the second largest retailer in the
world. Walmart is using Al-powered by data for improving the customer
experience at its site. It uses Al along with Internet of Things for clubbing
the online customer experience and in-store customer experience. For
Walmat
example, Scan and Go technology, Pick up towers, etc. Walmart is also
experimenting with facial recognition technology to see if its customers are
sad or hapPY while in the store or what within the store makes them sad or
happy.
Cortana: Microsoft Cortana is a virtual assistant that fulfills different roles. For
example, it can function as a chatbot on Skype or customer care executive or
provide news, weather, travel, etc., updates. Microsoft is also integrating Al
capabilities in its office productivicy suite. It is also providing third party access to
its Al platform for building the products.
Global Fishing Watch: Google combined the power of cloud
computing. machine learning. and geo-mapping using Al and
satellite data for building a system for preventing an illegal fishing. This Al system can predict
the reason for vessels to be at sea. Thisallows for a system that can identify instances of
illegal fishing. False positives can be verified by human law enforcementagencies GLOBAL
FISHING
WATCH

148 Kips I ARIFICIAL INTELLIGENCE

Google rain: Google rain project started as early as 2011. In the
Google restricted the use of this project to image recognition. With
starting, Goog.
Google started to use this Al for image enhancement. ow, Google
time

ine the same system for natural language processing and improving the
S de recommendations on YouTube by analysing the habits, behaviour, and
video
of the users.
preferences

Coogle is also using Al technology for developing self-driving vehicles.

THE EVER-CHANGING FOCUS OF AI RESEARCH

FIRST PHASE

The first phase of the Al research started with the Dartmouth Conference and involved techniques related
to the General Problem Solving (PS). During this phase, the researchers assumed that any problem could
be solved by using a program code and mathematical models. To solve these problems, the researchers
used computer data that was searched until a solution emerged.
For example, to solve chess, the whole problem was converted into a mathematical problems in which
the board. the pieces, moves, etc., were all converted into symbols. Searching through moves based on
the heuristics of previous games would allow the computer to find the best result. This approach showed
substantial initial benefits, but the approach did not continue for a long time.

SECOND PHASE
The GPS approach was not fit for solving real-world problems because the number of search
combinations increased exponentially with an increase in the problem complexity. The secon phase of
Al research focused on reducing the search space so that the searching can become better. During this
period:

attempts were made to find better ways of representing data.

the research in Al initially slowed down due to the negative publicity regarding them.
apan announced the Fifth Generation Computer Systems Project (FGCS) in 1982, which once again
increased the interest in the development of the Al systems.

THIRD PHASE
he third phase of Al research focussed on developing Intelligent Knowledge-Based Systems or IKBS.
h systems were also known as Expert Systems. These systems relied on specific domain-based data for
Ving Al problems. The main highlights of this phase were as follows:

IKBS worked on data in the form of rules.

KBS was restricted in the sense that it was incapable of learning. and its focus was extremely
narrow.
KES needed to be regularly updated.This was time-consuming and resource-intensive work.
BS
acked the common sense that made humans more suited for specific tasks and work.

FURTHER PHASES
The problem wit
oblem with Als is that all the research in the Al field is leading to the development of weak Als.
resea not made many breakthroughs in the development of the strong Als. The further phases of Al
involve developing systems that can learn and as such, would not need to be regularly updated.

ARTIFICIAL INTELLIGENCE Kips 149

 The ongoing developments in this area are:
Cyc project of Cycorp Inc. was the first project towards developing Al systems capable of common
sense
reasonings. The challenges of ambiguities of human language and high ambition caused the project
to slowdown.
The recent
attempts in this direction use a concept known as Big Data. Sometimes, this approach
involves capturing data from the web.
The development of techniques like neural networks and genetic algorithms for developing machine
learning
TYPES OF AI CAREERS
Al CAREERS CAN BE BROADLY DIVIDED
INTOTWOTYPES:
() Jobs requiring advanced training and education. These jobs usually are related to research,
development, and deployment of Al systems.
i) Jobs in which entry requirements are comparatively low.
of the Al systems.
These jobs are typically related to operations

Core Research Development

and Deployment Field and area spccialists

Software analysts Engineers (mechanical,

electrical, construction, etc.)

Software developers Medical professionals

Computer scientists Defence forces specialists

Computer engineers Craphic designers

Algoritham specialists Architects

Data scientists Musicians

Research scientists Entertainment specialists

Mochine learning engineers

Business intelligence
developer

Kips | ARTIFICIAL INTELLIGENCE

150
 CET REQUIRED FOR SOME AI RELATED CAREERS
SKILL SET R E Q U I

ine Learning Engineer

Machine Learniu One of the most in demand Al-related career. Responsible
Tor
developing and managing machine learning Projects
rstanding of different programming languages
nderstanding of applie research and data science
Strong mathematical skills

Data Scientists Individuals who love to play with data. Responsible for
collecting, analysing, and drawing interpretation from nuge
datasets of Big Data.
of Big Data platforms like Hive, Hadoop, MapReduce, etc.
Knowledge
Kowledge of statistical computing and programming languages, like Perl and Python
communication skills
Strong analytical and
Business Intelligence Subset of data scientists. Responsible for analysing data to
Developer identify business and market trends. Job requires design-
of
ing, modelling, analysing, interpreting and maintaining
datasets.

skills
Strong technical and analytical
Background in engineering, computer science,or related field
Experience of business and marketS
AI Research Scientists These scientists conduct research in the field of artificial
intelligence.They add to the knowledge of Al.
related field
Master or Doctorate degree in computer science or
Strong background in applied mathematics

Activity 2: Use of AlI Technology in a Real Life

ror this activity, match the company name in column | with the manner in which it is using artificial
intelligence mentioned in column l.

Column Column l
1. Deep Instinct A. Money laundering
2. PayPal B. Combating counterfeit products

3. Medecisions C. Cyber security

4. Burberry D. Improving logistics
5. Alibaba E. Improving healthcare

Solution: 4. B 5.-D
1. C 3. E
2. A
ARTIFICIAL INTELLIGENCE I Kips 151
Al ETHICS
ETHICAL CONCERNS RELATED TO AI ACCESS
Ethical concerns are one of the most critical problem areas that have emerged out of the
development of Artificial Intelligence and related technology. Broadly, we can divide these ethical
concerns into two
sub-problems-related to Data Management and the other, the adoption of Al
technolog8Y
CONCERNS RELATED TO DATA MANAGEMENT
Data is the lifeline of Al
systems. These systems cannot exist without the data. But the collection, storing,
and usage of this data raises serious
privacy concerns. Take a simple example of Google, which uses data of
its users for
increasing its revenue. One of the products offered by Google is the Android operating system.
This operating system comes with a number of
have provided permissions in some
google apps pre-installed. So if you are using these apps and
instances, then Google has the following information related to you:
1. Contact List:
Google knows the name and phone number of all of your friends, relatives, business
partners, colleagues, etc. from your phone book. Also, it know the names of the persons with
whom
you like to talk and with whom you do not like to talk, and the best and worst time to call
2. Location: you.
Google knows the location of your house, your friend's house, houses of your relatives,
your school, office, favourite mall, cinema house etc. It also knows the time
you spend in these
locations. the routes you take to reach them. With a little effort it can
predict you own a
if
vehicle,
usepublic transportation system, carpool or your parent's chauffeurs you.
3. E-mail: Google knows every
single mail written and received by you. It also knows about the mails
that you never read and the mails that you wrote but never sent. It knows about
the job offers
received by you, and about the job you were
rejected from.
4. Hangout: Google knows every little secret shared by or with you. It also knows all about
friends, their preferences, personal life, etc. (which was shared in your
5. Photos: All the images of you and your friends
hangout chats).
along with their names linked with the contact list;
Google knows their phone numbers, email-IDs, addresses, etc. Linked with calendar, it also knows
about important events like their
birthdays, marriage anniversaries, birth of their children, etc. Most
of the time, we accept lengthy user
agreements without even realising the implications of these
agreements on our privacy rights. Data related to the users is stored somewhere in some
databases. This data can be used for a number of large
purposes.
The information is collected in lieu of
providing us with helpful services, but the potential risk here is very
high. For example, take the facial recognition system. The Al system can use this technology along with
the information provided by Apps like Contact and
Photos, to create an extensive surveillance network
that can monitor you 24x7.
This is just a small amount of your personal data, which
Google has. Think about all of the details regarding
you with Google, i.e., your search history, internet browsing habits, phone using habits, etc. Add to this
the data that other companies like Microsoft, Facebook, Twitter,
Amazon, etc. use. Try to imagine the
extent and the depth of the data that these companies have, and you will have a clear understanding of the
ethical concerns related to data collection.

BIAS IN REAL-WORLD DATA

The problem here is that the Al system learns from the real-world data fed into it. This means that Al
systems can reinforce the biases found in the Al systems. For example, a computer system trained on the

152 KiPs I ARTIFICIAL INTELLIGENCE

 200 years might find that more females were involved in specific jobs or that more percentaE
last quipped
better egu
data 1o l businesses were established by men and conclude that specific genders are
o fs u c c e s s f u l

certain
jobs
job (gender bias
handling

for black boxes.

detecting such biases is not easy because Al systems act as
ding or even
many
derstand. Many
Understa hehind their decision-making is not easy or in some cases, even possible to under
by the Al
explain the logic behind decisions taken by
The reasc

themselves
times
orammers of Al systems cannot

systems.

INCLUSION
PROBLEM OF
THE oblem that
A ls y s t e m s t r a i n e d on biased real-world data create the of inclusion, i.e.
problem
the Al
the pro
Al decision-making system. Consider the example of system
neOple are left out of the females were left out of
eligible
some

This created a situation in which many

by Amazon for recruitment.
Used
nsideration. This is known as the problem of inclusion.

FACTS ANDTHEIR INTERPRETATION

THE PROBLEM OF
of the bias in Al systems from the problem between facts and the interpretation
come
The root cores to
can scan data and draw learnings from them, but Al systems are not equipped
of facts. Al systems
behind a particular conclusion o r learning.
understand the reason

experimental twitter chatbot known as Tay: Tay was designed

to
Microsoft developed and launched an

conversational understanding. VWhile launching Tay, Microsoft

claimed that the m o r e yoou
experiment with conversations.
to engage people through casual and playful
chat with Tay, the smarter it will get, learning
learned from thousands of tweets and developed a misogynistic
In less than 24 hours of its launch, Tay
racial personality.
The problem was that Tay did exactly
There was no programming e r r o r o r hacking for the consequences.
who were talking to it. lt learned from what
what it was designed to do. lt started mimicking the users
was being fed to it without interpreting the inputs.

cOMPONENTS OF A GOOD AI POLIcY

Good Al Policy

Al must have the right 1o he data is collacing

it
An

must have a choice of leaving the

Al system
Consumers

dafa collected and the purpose of the Al must be mited by desion

The
request from the consumers
Data with the Al system mUst be deletec upon

HICAL CONCERNS RELATED TO ADOPTION OF THE AI SYSTEMS

The second
cond set
set of ethical concerns are related to the adoption of the Al systems. Some of the critical
ethical concerns elated to the adoption of the Al systems are:

ARTIFICIAL INTELLIGENCE |
Kips 153
JOB LOSs
The argument on the
limplications of the adoption of the Al system has been going on since the introduction
of Al. There is no consensus on
the issue. Some argue that Al
it will increase systems will reduce job; other argue that
jobs. There are still some who claim that there will be no effect on the
following studies conducted by some of the reputed international firms, and jobs., Consider the
will better
you understand
this issue
McKinsey Global Institute: This institution has
predicted that almost 40 million to 160 million women
are
going to be affected across the world due to the adoption of Al
clerical, secretarial, schedulers, systems. They claim that jobs like
such jobs are held by women in the
bookkeepers etc. are highly
susceptible to automation and almost 72% of
developed economies alone.
Oxford Economics: Studied the
robots and concluded that
manufacturing sector and the world for jobs which can be replaced by
adoption of Al systems approximately 20 million manufacturing jobs will be lost
by 2030.
McKinsey Global Institute: In
one of the studies on the
issue, the institution has predicted that in
countries like che USA some areas can see
rural areas can see more job loss
job loss of 25%, and other areas can see
job loss of 20%. The
compared to urban areas.
World Economic Forum: They studied the issue and said
that by 2022, Al would lead to job loss of 75
million. But they claim that this will be offset
during the same period.
by the 133 million new Al-related
jobs which will be created

INCREASING INEQUALITIESs
The current economic
system provides economic rewards based on contributions to the economy. This
is true for all the economies, i.e.
capitalist, socialist, communist, mixed, etc. The direct consequence of Al
adoption will be reduction in number of people required for doing certain
result in transfer of economic benefits from the specific jobs. This, in turn, will
persons doing these jobs to the companies managing the
Al system. This will increase the income and lead to
gaps concentration of economic benefit in the hands
of the few individuals.

NEGATIVE ADOPTIONS
History shows that human beings are capable of finding negative or evll uses of every technology ever
discovered. This is also true for Al systems. In the hands of the terrorists it can
become a
potent tool of spreading terror. In the hands of cybercriminals this can be a dream come truepotentially
for their
hacking and damaging needs, and in the hand of rival countries these technologies can spell doom for us.

BLACK BOx PROBLEM

Al systems especially the ones based on neural networks
cannot explain the reasoning which
they used
-

for arriving at a particular decision or solution. Most of the

time, even their creators cannot explain the
process of this decision making. This is known as the Black Box problem. What happens insider the Black
Box remains in the Black Box forever.

COST BENEFIT ANALYSIS OF THE AI SYSTEMS

The main ethical question before us is to evaluate these benefits against the human costs to see if, in
the specific case this adoption is justified or not. We must not
forget that most of the technologlcal
advancements of past have also with limitations
come or dangers.

154 Kips I ARTIFICIAL INTELLIGENCE

a Solution does not lie
with throwing away
the
limitations and rea educing the
dangers. technology, but it lies in finding the way of
Every technological
ind as well as well as its destruction.
mankin
ve can use development can overcoming
be used for the benefits of
weapons. Nuclear power can be used to
them as weapons. dynamites to make mining easy, and we can an also use
create energy, and it
and countries. Advancement in medicines can be used
can also be used to destroy cities
to heal and
heings as well as to create
biological weapons. Al improve the quality of life of
human
mankind technology is
different. It can be used for the good ot
no
man
Al cvstems
or its
destruction.
as it has done for
Overal, the society will have
to come
up with ethical
many of the other
technologies. The violations guidelines
on data anda
of these
regulations will have to be punished appropriately guidelines, rules or

WORKSHEET 6: ldentifying Ethical and Privacy Concerns

For this activity, open the smart phone of someone in your family.
them first!). List out all the aPps on the phone in the table below.
(Take permission fromn
Identify the privacy concern
associated with each of these apps. Write, 'No Concern', if you think that the app does not
have any privacy concerns. (Do not use apps mentioned in the first part of this section)
AppName Privacy Concern()

ARTIFICIAL INTELLIGENCE | Kips 155