Professional Documents
Culture Documents
Artificial Intelligence
Artificial Intelligence
Artificial Intelligence
Basic Guide to AI
Sandeep Kulshrestha
SANDEEP KULSHRESTHA
ARTIFICIAL
INTELLIGENCE
BASIC GUIDE TO AI
2
Artificial Intelligence: Basic Guide to AI
1st edition
© 2021 Sandeep Kulshrestha & bookboon.com
ISBN 978-87-403-3885-0
3
ARTIFICIAL INTELLIGENCE Contents
CONTENTS
Preface 5
4
ARTIFICIAL INTELLIGENCE Preface
PREFACE
This book on Artificial Intelligence is a basic primer meant to help you understand the
subject in a very simple and concise manner. As a writer of this book, I have kept in mind
to explain each and every thought, concept or idea in a very straightforward way that I am
sure would be helpful. If you are keen on exploring the practical side of AI programming
and coding, I suggest that you can start with free or low-priced courses on Google, Coursera,
edX and other such platforms. These days many people also offer basic coding techniques
on private videos (on YouTube) as well.
So, enjoy this journey in getting to know Artificial Intelligence. I hope this small book would
channelise your thinking and make you imagine what the future world looks like, with AI.
Sandeep Kulshrestha
Hyderabad, India
5
ARTIFICIAL INTELLIGENCE Introduction to Artificial Intelligence
1 INTRODUCTION TO
ARTIFICIAL INTELLIGENCE
We are living in the world of “Alexa” and “Siri”, and we know what type of work they
are capable of. We are told that they are “AI” based applications that work efficiently on
voice commands. Yes, indeed they are based on something that we hear much often these
days – Artificial Intelligence.
Let us hence begin our explorative journey of the term, “Artificial Intelligence” by looking
at the two words separately. If you take a few seconds pause, what does the word “artificial”
signify to you? What all things or objects that you imagine are artificial? Perhaps your study
table, or the plastic flowers you see sometimes at a restaurant or any toy, that you had
played with in your childhood. Isn’t it? Can we say except for the living objects, the flora
and fauna and the universe, everything that is man-made is artificial? Oxford dictionary says
this about artificial: made or produced to copy something natural. Like a soft toy is mimicked
after a real animal and hence is artificial.
Now let us look at another word, Intelligence, which means the ability to acquire knowledge
and the collection of information which would be useful for any current or future needs.
Why do we acquire knowledge? To apply somewhere, whether it is in a business or in a job.
So, when you apply for a job, the recruiters look for your experience and your intelligence.
Hence, when we combine two words, Artificial Intelligence, it signifies creation of man-
made inanimate objects (let us call them computers or devices powered by computers) which
imitate or copy human intelligence. For example, let us imagine that everyday evening at
5 pm, you need to sit at your study table. It is an everyday affair, except for the weekend
where you need the table at 10 am. If it is an artificially intelligent table, it will come to
you at 5 pm on weekdays and at 10 am on weekends because it has become intelligent
(just like a robot)!
Hence, when machines start getting their own intelligence and they learn all the time, they
are artificially intelligent machines.
Activity
6
ARTIFICIAL INTELLIGENCE Introduction to Artificial Intelligence
AI is often compared to human performance, and we will often take humans as inspiration
for how computers might solve similar problems.
One of the most important goal of artificial intelligence is to create expert systems, which
exhibit intelligent behaviour, continuously learn on their own, demonstrate, explain, and advice
its users. Like you have experts in their respective fields who make human understanding
easier, similarly the machines act as experts to make our lives easier. For example, when you
ask Alexa or Siri to play a recently popular song, as an expert who is continuously learning,
AI enabled gadget will play you a song. This is of course a very simple example of AI.
Here, the idea of Artificial intelligence is to Creating systems that understand, think, learn,
and behave like humans. The example is robots which are mostly used in manufacturing
units. The dependency on robots to undertake human ventures is also commonly termed
as “automation”. Besides the usage of robots in factories, the smaller version of robots has
increasingly been used at homes, in the forms of gadgets like automated vacuum cleaner.
Pictures of a robot and an automated vacuum cleaner are provided here:
7
ARTIFICIAL INTELLIGENCE Introduction to Artificial Intelligence
8
ARTIFICIAL INTELLIGENCE Introduction to Artificial Intelligence
1.3 APPLICATIONS OF AI
AI is already applied in various fields and some of them are mentioned here;
Gaming
AI plays a critical role in strategic thinking related games such as chess, poker, tic-tac-toe,
etc., where machine can think of large number of possible positions based on heuristic
knowledge.
Expert Systems
Expert systems are created by the merger of machines, software, and continuous learning to
impart reasoning and advising. Such systems have the capabilities to provide explanation and
advice to the end users. Let us imagine that you are a hiring manager, and your company has
created an expert system for screening the candidates. In this context, your task is reduced
only to meet the candidates as the expert system has already filtered the candidates. This is
fairly a basic example, and we will look at other examples as well.
Vision Systems
These systems understand, interpret, and comprehend visual input on the computer.
For example, A spying aircraft takes photographs which are used to figure out spatial
information or map of the areas. Doctors may use clinical expert system to diagnose the
patient. On the other hand, the Police uses computer software that can recognise the face
of criminal with the stored portrait made by forensic artist. You must have seen few of
these things in the movies.
9
ARTIFICIAL INTELLIGENCE Introduction to Artificial Intelligence
Speech Recognition
AI enables intelligent systems that are capable of hearing, understanding, and comprehending
the language in terms of sentences and their meanings while a human talk to it. It can
handle different accents, slang words, noise in the background, change in human’s noise
due to flu and so on
Handwriting Recognition
The handwriting recognition software reads the text written on paper by a pen or on screen
by a stylus. It can recognize the shapes of the letters and converts it into an editable text.
This is very common.
Intelligent Robots
Robots can perform the tasks given by a human. They have sensors to detect physical data
from the real world such as light, heat, temperature, movement, sound, bump, and pressure.
They have efficient processors, multiple sensors and huge memory, to exhibit intelligence.
In addition, they are capable of learning from their mistakes, and they can adapt to the
new environment.
1. Reactive Artificial Intelligence: This is the most basic of AI systems where the
machine does not have memory and it reacts. For example, when a computer
system defeated a chess player, it was sort of a Reactive Machine. It only knew
how to play the game.
2. Limited Memory: These systems or machines are reactive machines and in
addition, they have historical data, helping them take decisions.
3. Theory of Mind Artificial Intelligence: This is an advanced AI technology that
seeks to understand human behavior, intelligence, emotions, perceptions
etcetera. One example of such AI is “Kismet”, designed in late 1990s. Kismet
can mimic human emotions and also recognise them.
10
ARTIFICIAL INTELLIGENCE Introduction to Artificial Intelligence
4. Self-Aware AI: As the name suggest, these machines are self-aware, as they
show in the movies, where the objects mostly from outer space (and look like
machines) come and destroy the population and habitat. It is still in nascent
stage but many scientists want such machines which would be self aware
and of course would not be out, to destroy but to help humans in making
their lives better.
Activity
5. Artificial Narrow intelligence: It is also known as “Weak AI” and this is actually
the most frequently used AI in the world around us. Voice recognition on your
mobile applications like Google, Siri or Alexa is an example of weak AI as they
recognize your voice based on the dataset, they are trained in.
6. Artificial General Intelligence: Also known as “strong AI”, allows a
machine to apply knowledge and skills in different contexts. This
resembles more like mimicking human intelligence through autonomous
learning and problem solving.
7. Artificial Super Intelligence: This is something more advanced than even the
strong AI. These would be way better than human intelligence and the pitfall is
that it may well end up being destructive.
11
ARTIFICIAL INTELLIGENCE Introduction to Artificial Intelligence
certainly is. The Google’s AI structure looks at what you have been watching.
You follow some steps. First you login to YouTube, then type in what you
would want to see and then you choose to click on the video. The AI learns
from your frequent views that you like few types of videos and it would start
recommending you more of such content based on your choices. An algorithm
does that. It learns from the user’s data.
b. Natural Language Processing: This was mentioned earlier in the chapter as well.
NLP refers to the processing of human language by a computer with a purpose
of adding value to AI
c. Machine Learning: In simple words, Machine learning is like human learning,
but in a different context. For example, you know the routes of reaching your
friend’s house and know which one would take less time, to reach. In other
words, you have the data of these routes based on your experience. Similarly,
through Machine Learning, an AI system understands data, learns from it and
help in making decisions that add value
d. Deep Learning: Deep learning is a sub-set of Machine learning and as the
name suggest, it is surely deep. Through deep learning the machines can learn
without supervision, from the data that is either unstructured or unlabelled.
In simple terms, let us say that you go to various people’s houses in a day and
randomly change routes. Through deep learning the computer would be able to
predict your movement of next weeks. This is a basic example of deep learning.
e. Coding (if you are new to computer programming): Coding is the new term
for computer programming. Through coding we communicate to the computer.
All the wonderful wallpapers you see, or a beautiful Microsoft document
you create have emerged out of people making codes, to give instructions to
the computer. You can have fun through Coding and can create wonderful
websites, games and many other interesting stuff. You can go to https://www.
codecademy.com/learn and learn coding for free
f. Python: Python is a programming language that is useful in creating computer
programmes for all purposes and can be used for making websites or even
creating and developing AI applications. There are many instructors on udemy.
com and on YouTube who offer courses for beginners
12
ARTIFICIAL INTELLIGENCE History and discoveries in Artificial Intelligence
So, by the 1950s, there was a worldview of Artificial Intelligence in the minds of scientists,
mathematicians, and philosophers, maybe in not the same understanding that we have now.
One of such creative souls was Alan Turing who was a young Britisher (and a polymath!)
and he looked at the possibility of Artificial Intelligence, through the lens of Mathematics.
Turing’s logic was if humans could use their expertise and resources in solving problems
and make decisions, why couldn’t the machines do the same things? In his 1950 paper,
Computing Machinery, and Intelligence, he deliberated on building Intelligent machines
and how we could test and utilise the intelligence of such machines
13
ARTIFICIAL INTELLIGENCE History and discoveries in Artificial Intelligence
The second challenge was that the computers were overly expensive. So much so, the cost
of leasing out a computer was as high as $200,000 which could only be affordable by big
corporations and Ivy league Universities. Also, if one wanted to attract funding to produce
superior machines, one needed a proof of the concept and advocacy from high profile people.
In that context, it was very difficult to work around future technologies.
14
ARTIFICIAL INTELLIGENCE History and discoveries in Artificial Intelligence
Mr. McCarthy was very passionate about the cause of Artificial Intelligence and he brought
together imaginative researchers who spoke on an open platform. In fact, the term “Artificial
Intelligence” was coined at the same conference by McCarthy himself
McCarthy had great expectations from the conference. Unfortunately, people came and went
and could not agree on any standard methods for the upcoming field. As any event has
few positive take-aways as well, many of the participants of the conference agreed with the
thought that AI was achievable. This event led to next two decades of interesting research
in the area of AI, hence we cannot really tone down the importance of the conference.
Now, when anyone says the word “Dartmouth”, you will most probably remember it for
its association with AI
These successes, as well as the advocacy of leading researchers (namely the attendees of the
Dartmouth conference convinced government agencies such as the Defense Advanced Research
Projects Agency (DARPA) to fund AI research at several institutions. The expectations
from the technology were at its peak and so was the levels of optimism. There were several
challenges and the prominent one was lack of computational power to achieve any substantial
amount of results. The information storage capacity of computers was not at all good
enough. A man called Hans Moravec, then a PhD student under McCarthy (McCarthy
had stints of Professorship in Computer science at Stanford, MIT and Dartmouth) stated
once, “computers are still millions of times too weak to exhibit intelligence”
15
ARTIFICIAL INTELLIGENCE History and discoveries in Artificial Intelligence
introduced expert systems which mimicked the decision-making process of an expert Human
being. So, essentially the AI program would ask an expert in a field how to respond in
different situations, and once this was learned for virtually every situation, non-experts could
receive advice from that program. Expert system hence became popular in many industries.
As part of their Fifth Generation Computer Project (FGCP). From 1982-1990, the Japanese
government heavily funded expert systems and other AI related endeavors and they pumped
in an investment of $400 million dollars with the goals of revolutionising and improving
artificial intelligence. Although this program did not kick-off as expected but at least it
brought in enough interest in the computing and scientific fraternity.
Although there was apathy in terms of funding by different Governments, still AI thrived
and during the 1990s and 2000s, much was achieved in AI. Famously, IBM’s Deep Blue,
a chess playing computer programme defeated World chess champion Gary Kasparov in
1997. This event was nevertheless highly publicised as it was the very first time that a world
grand master lost to a computer programme. In a sense, it was a big deal for AI as this was
a giant leap towards developing more such programmes that could learn, understand and
make decisions. The same year, 1997 saw another breakthrough when Microsoft platform
introduced a speech recognition software, which was developed by a company called Dragon
systems. The time was exciting for the researchers of AI and the computing world, and
the experts started believing that there was nothing that the machine could not do. MIT
scientist Cynthia Breazeal developed a “social robot” called Jibo that could display and
understand emotions. You can hear Cynthia’s TED talk at this link: https://www.youtube.
com/watch?v=eAnHjuTQF3M
Activity
16
ARTIFICIAL INTELLIGENCE Understanding Human Intelligences
3 UNDERSTANDING HUMAN
INTELLIGENCES
Before delving into the finer details on Artificial Intelligence, we need to also look at various
human intelligences that make us unique. When you hear the term, “intelligence”, what
comes to your mind? In this chapter we will look at varied human intelligences and we
would ponder if the machines can mimic the intelligences we possess as a human. Now
as a reader, when you read this book, you have a certain intelligence within you to either
understand the ideas or concepts fully or partially. Can this book be converted into an AI
app? There is surely a possibility. Or can all the human intelligences be automated by a
computer? If not all, than would it be few that could be converted to an AI framework?
While you read forward, perhaps you can use your own thinking and intuition to visualise
what all intelligences can be converted to AI
• Learning and Reasoning ability: When we can process new information and
are able to apply that information for taking a decision, we are supposed to be
using the intelligence to learn, reason and apply.
• Ability to comprehend and grasp facts: Human beings have this ability to
understand. Even as a kid we understand when our parents are angry or what is
being taught at the school
• Understanding personal relationships: We humans have the capability of
understanding and maintaining personal relationships in terms of making
opinion about other people, understanding other people’s strengths and
weaknesses, evaluating people critically and also mending broken relationships
by being empathetic and compassionate. Surely we cannot imagine a machine
understanding and managing human emotions
17
ARTIFICIAL INTELLIGENCE Understanding Human Intelligences
• Ability to sifts facts and beliefs: As humans, majority of us have the ability to
make a difference between facts and fictional beliefs and that is how there are
very few takers for conspiracy theories or something weird we hear that makes us
think, “nah, this doesn’t look to be true”
The author has also tried to categorise human intelligences, while bearing in mind if the
following intelligences could be mimiced by computer enabled machines, with a purpose of
aiding and helping human beings developing innovative solutions to their specific problems
18
ARTIFICIAL INTELLIGENCE Understanding Human Intelligences
Activity
Activity
19
ARTIFICIAL INTELLIGENCE Understanding Human Intelligences
Creative Intelligence
Spatial/Geographical
Intelligence
Kinesthetic Intelligence
The intelligence
of interpersonal
communication
Linguistic Intelligence
Mathematical or Logical
Intelligence
Even if you did not attempt the above activity, your mind will continue working on various
things that the AI can do. You have picked this book for a reason. It may be either based
on curiosity, academic interest, or a future plan of action. Whatever the reason may be,
you will be able to use your imagination, as a stimulus to analyse what all can be achieved
through AI. One example of an application that works in the area of wellbeing, is BioBase,
which is developed by a company called BioBeats. The app is used mostly be enterprises
who collect data from wearable device called BioBeam. The data collected helps monitor
well-being and facilitates providing of real time insights and feedback. The use of this App
has helped reduce employee absenteeism by a substantial percentage. Hence, we can use AI
in a lot many ways, even when it comes to areas like mental wellbeing.
Human intelligence has created Artificial Intelligence. But would the AI become better than
humans? That does not look likely as of now. Future surely may surprise us though. When
you read the philosophy of AI in one of the chapters in this book, you may think more
on this aspect. Of course the future surely belongs to AI.
20
ARTIFICIAL INTELLIGENCE Introduction to Machine Learning
4 INTRODUCTION TO
MACHINE LEARNING
Machine learning, as a term was coined in 1959, by Arthur Samuel, an American innovator
in the field of Artificial Intelligence and computer related gaming. Machine learning implies
that computers could learn on their own without being specifically programmed. In 1997, the
American computer scientist Tom Mitchell said, in mathematical terms, about a computer
program that learns, “A computer program is said to learn from experience E with respect
to some task T and some performance measure P, if its performance on T, as measured by
P, improves with experience E”
Now that data science has emerged as a field we cannot live without, Machine learning has
become an important tool, used to create complex algorithms and models that can finally
predict the outcomes. Let us imagine a data scientist gets big data about all the customers
who wore more fashionable outfits at any of the high-end shopping mall (Like H&M)
and their spending pattern. Through Machine learning we can predict that customers with
certain preferences would end up buying a specific kind of new clothing line. This is again
a simple example. Much more complex programs are being created for Machine learning
in the world we live in and it may become even more interesting in the future.Let us look
at another example. Let us imagine that you decide to check out an offer for a vacation.
You browse through website like booking.com and search for a hotel stay, say at Tahiti.
When you look at a specific hotel, just below the hotel description there is a section titled
“You might also like these hotels”. This is a common example of Machine Learning and
is referred as “Recommendation Engine”. Again, many data points were used to train a
model to predict what will be the best hotels to show you under that section, based on a
lot of information they already know about you. Forget about booking hotels for a while.
If you visit travel apps on your mobile phone more often, you may just look around for
good hotel properties, just for research and not necessarily for vacation. The AI predictor
would still understand your choices and recommend you other hotels.
21
ARTIFICIAL INTELLIGENCE Introduction to Machine Learning
If you want your program to predict, for example, traffic patterns at a busy Traffic intersection
(task T), you can run it through a machine learning algorithm with data about past traffic
patterns (experience E) and, if it has successfully “learned”, it will then do better at predicting
future traffic patterns (performance measure P).
The highly complex nature of many real-world problems, though, often means that inventing
specialised algorithms that will solve them perfectly every time is impractical, if not impossible.
Examples of machine learning problems include, “Is this cancer?”, “Which of these people
are good friends with each other?”, “Will this person like this movie?” such problems are
excellent targets for Machine Learning, and in fact, machine learning has been applied to
such problems with great success.
As a kind of learning, it resembles the methods humans use to figure out that
certain objects or events are from the same class, such as by observing the degree
of similarity between objects. Some recommendation systems that you find on
the web in the form of marketing automation are based on this type of learning.
22
ARTIFICIAL INTELLIGENCE Introduction to Machine Learning
Errors help you learn because they have a penalty added (cost, loss of time,
regret, pain, and so on), teaching you that a certain course of action is less likely
to succeed than others. An interesting example of reinforcement learning occurs
when computers learn to play video games by themselves.
Hence, Machine learning can be classified into various subsets. To understand Machine
Learning in a simple manner, you can see a video here: https://www.youtube.com/
watch?v=VwVg9jCtqaU
23
ARTIFICIAL INTELLIGENCE Robotics and Artificial Intelligence
AI technology opens whole new vistas of economic opportunity by enabling robots to work
where people cannot. Robots are of great value for all sorts of tasks that are too dangerous
or costly for people to do. These might be mining or farming the sea bottom, eliminating
agricultural pests by targeting them with insect-specific mechanical predators, or cleaning
up industrial accidents.
One obvious such area is exploration of space. In 1993, NASA sent the space shuttle with
seven people on board on a mission to repair the Hubble space station; the objective was to
perform an exceptionally precise operation to correct the space telescope’s optics. This was
the first of five manned Hubble maintenance missions. In 2004, serious consideration was
given to using a two-armed Canadian robot called Dextre instead of astronauts for the final
mission, but it was judged too risky. Nonetheless, robotic devices are likely to be much more
practical for the sorts of tasks we are likely to want to perform perhaps on other planets in
the near future, such as analysing geological samples, searching for biological life, mining
asteroids, and diverting astronomical bodies whose paths threaten earth.
24
ARTIFICIAL INTELLIGENCE Robotics and Artificial Intelligence
There are robots for entertainment. These usually take anthropomorphic forms like the
preprogrammed animatronics figures common in theme parks such as Disneyland but are
considerably more flexible and interactive. Pepper, from Alderbaran Robotics and SoftBank
Mobile, tries to read your intentions and respond appropriately. It is currently used to greet
visitors to SoftBank stores in Japan, where it can answer limited questions about products
and services, but its primary value is to engage and delight customers. There have also been
many generations of interactive toy robots.
One of the most exciting recent developments in the field is own as “swarm robotics.” Large
collections of relatively simple uniform robots are programmed with rules, and when these
are applied in aggregate to the entire group, the robots exhibit complex behavior, called
“emergent behavior.” TIUs same effect is observed in anthills and beehives, whose members
as communities solve problems that are far beyond the comprehension or capabilities of
any individual. While swarm robots could be any size, much
It’s difficult to exaggerate the potential benefits and dangers of this technology. On the
positive side, it could facilitate tremendous medical advances, such as performing noninvasive
surgical procedures from inside the body. Imagine a syringe full of robots the size of T-cells
that mimic the function of the immune system, able to seek and attack blood home cancers.
Or a shoebox full of robots the size of cockroaches that scurry around collecting dust from
floors and walls, stuffing their bounty into a small bag for easy disposal. Imagine releasing
thousands of mole-sized robots to explore for minerals underground, followed by tiny
robotic miners.
25
ARTIFICIAL INTELLIGENCE Robotics and Artificial Intelligence
But there are also significant dangers. The same technology that might cure blood borne
cancers can be used to kill you, or perhaps even to control you.” Anyone who has tried to
eliminate ants from the kitchen knows how difficult it can be to prevent an invasion by
tiny, organised army. The potential military or terrorist applications of swarm robotics are
truly too horrific to contemplate.
Military applications are too numerous, and perhaps dangerous, to mention. While the
popular imagination conjures up visions of Terminator-style robotic soldiers running around
a theater of battle bearing arms, the truth is very different. Military robots will not be
designed to use weapons, they are the weapons. Examples include guns that can identify
targets and shoot autonomously, flying drones that can deliver explosive charges to precise
locations, and land mines that explode only when specific enemy vehicles are within range.
The possibilities are so disturbing those significant efforts are under way by the United
Nations and the military establishment to study the ethics and efficacy of using such precise
munitions to support or replace personnel in war zones. The current consensus is that as
a matter of caution, a human should be “in the loop” for all targeting decisions before
pulling the trigger, but it’s not entirely clear that this is practical, or ethically defensible,
since requiring such review may put lives at risk.
26
ARTIFICIAL INTELLIGENCE Philosophy of Artificial Intelligence
6 PHILOSOPHY OF ARTIFICIAL
INTELLIGENCE
Artificial Intelligence does pose real challenges to philosophical and religious doctrine about
human uniqueness and our place in the universe. Intelligent machines offer the potential
to shine an objective light on fundamental questions about the nature of our minds, the
existence of free will, and whether nonbiological agents can be said to be alive.
The philosophy of AI asks the question of whether computers, machines in general, or for
that matter anything that is not of natural origin can be said to have a mind, and / or to
think. The answer, simply put, depends on what you mean by “mind” and “think.” The debate
has raged on in various forms-unabated and unresolved-for decades, with no end in sight.
The use of symbols for reasoning and communication. A distinction is commonly made
between syntax, the rules for arranging and manipulating symbols, and semantics, the meaning
of the symbols and rules. While syntax is easy is to understand, semantics is not-even the
experts don’t agree on what “meaning” means. Most theories propose that meaning requires
some way of relating the symbols themselves to the things they denote in the real world.
27
ARTIFICIAL INTELLIGENCE Philosophy of Artificial Intelligence
1+1=2
1+2=3
2+1=3
l+3=4
3+1=4
2+2=4
We all know what 1, 2, 3 and 4 mean, except for the minor inconvenience that they don’t
actually mean anything more or less than !, @, #, and $ do. They derive their meaning
from how we connect them to other concepts or real-world objects. If we connect $ with
any collection of four things, an expanded set of the above rules is exceedingly useful for
solving certain problems of great practical significance.
28
ARTIFICIAL INTELLIGENCE Philosophy of Artificial Intelligence
And to do something requires some connection between the actor manipulating the symbol
system and something external to that actor. In the case of computer programs, this could
(for instance) be figuring out how much you owe the phone company this month, the
movement of a chess piece (physically or virtually), or a robot picking up a pencil. Only
in this context can you say that the symbol manipulations have meaning.
Now, common arithmetic is one thing, but a vastly expanded concept of symbols and
rules is a reasonable description of just about any computer program at some level, even
if it’s possible to make other interpretations of those same programs. It’s an incredible eye-
opener for most computer science majors when they first discover that all the math they
ever learned in high school is simply a special case of some surprisingly easy to understand
more general rules.
Computers, by themselves, cannot “think” in this sense at all, since they don’t actually
mean or do anything-at best, they manipulate symbols. We are the ones associating their
computations with the external world. Electrons may be floating around in circuits, but we
are the ones interpreting this activity as symbol manipulation.
While the symbols themselves may be devoid of any semantics, perhaps the meaning arises
out of their relationships to other symbols, just as the definition of a word in a dictionary
is expressed in terms of other words. Machine learning algorithms suffer from the same
conceptual (though not practical) shortcoming-they reflect the complexity of the real world,
but without some connection to that world, it is just so much unmoored structure.
What is the difference between ideas swirling around in your brain and bytes zipping
around in a computer? In both cases, information is going in, represented in some form
that can plausibly be called symbolic (discrete nerve signals from your eyes, for example),
getting processed, and coming back out (nerve signals to your hand to press keys on your
keyboard, resulting in a spreadsheet of total monthly sales).
These must, in fact, be different things, but we just do not understand yet what the brain
is doing. The program is simulating thinking, not duplicating the process that occurs in
human minds when they engage in these activities.
Computer programs, taken by themselves, do not really square with our commonsense
intuition about what it means to think. They are “simply” carrying out logical, deterministic
sequences of actions, no matter how complex, changing their internal configurations from
one state to another. But here’s where we get into trouble: if you believe that our brains
are little more than symbol manipulators composed of biological material, then you are
naturally forced to conclude that your brain, by itself, can’t think either. Disconnect it from
29
ARTIFICIAL INTELLIGENCE Philosophy of Artificial Intelligence
the outside world, and it would be doing just what a computer does. But that does not
square with our commonsense intuition that even if we sit in a dark, quiet room, deprived
of all input and output, we can still sit there and think. We cannot have it both ways: if
symbol manipulation is the basis of intelligence, either both people and machines can think
(in principle, if not in practice today), or neither can.
But if you prefer to maintain the comforting conceit that we are special-different from
machines in some fundamental way yet to be determined or that we are imbued with some
mystical qualities quite distinct from the rest of the natural world, then you can cling to
the notion that thinking is uniquely human, and machines are simply pretenders to our
cognitive abilities. It’s your choice. But before you make up your mind, bear in mind that
there’s an accumulating body of evidence chipping away at our seemingly obvious intuitions
about our most quintessentially human abilities-for example, that we actually have free will.
Assume that computers cannot have free will because they are different from us in two key
respects. First, they work according to well-understood engineering principles and so can
always be predicted. Second, they cannot really be said to consider choices in the same
sense that people do. The problem is that both of these assertions are questionable at best.
The physical world operates in accordance with certain laws of nature, whether we know or
can know what those laws are. This is not to say that everything is predetermined indeed,
randomness may in fact be a fundamental part of nature. But randomness is just that random,
not a free pass for things to happen in accordance with some grander plan or principle that
is somehow outside of the laws of nature. Otherwise, those plans would simply be part of
the laws. In other words, there is no such thing as magic. Assume that your mind arises
from your brain, and your brain is a physical object subject to the laws of nature.
30
ARTIFICIAL INTELLIGENCE Philosophy of Artificial Intelligence
Now imagine that we can build a robot that does exactly what you can do. No matter
how we programme that robot to make decisions, no matter how predictable that robot
is, if it has access to an outside forecast of its own actions, that forecast cannot always be
correct. The robot can simply wait for that forecast, then do the opposite. So, a sufficiently
capable robot can’t always be predicted, where “sufficiently capable” means it has access to
the attempt to predict what it will do.
A deterministic machine, whose behavior is completely specified and understood, can always
be predicted. Any given state of a computer program may transition to its next state in an
entirely predictable way, but surprisingly, we cannot simply string knowledge of these states
together to get a complete picture of what the program will ultimately do. And the same,
of course, is true for you, you can never accurately predict your own behavior. It is possible
that therefore we have the strong intuition that we have free will, but this is simply an
interesting hypothesis, not a proven fact. Other possibilities are that our subjective sense of
free will has arisen to serve some yet to be identified evolutionary purpose(s), like desiring
sweets or being attracted to the opposite sex.
Look at the question of how computers make decisions. Using metaphor and analogy to
solve the problems. Our sense of choice is nothing more than an illusion. Presumably, your
brain, as a physical object, plays by the same rules as the rest of the physical world, and
so may be subject to inspection and analysis. And if your mind arises from your brain, at
some level it too must operate according to some laws of nature, whether we understand
those laws yet or not. Introducing randomness into the picture does not get around this
problem, and neither does the peculiar fact that lots of deterministic processes are nonetheless
not subject to prediction even in principle. Finally, there is no reason other than wishful
thinking to suggest that machines are in this regard any different from us. This is not to
say that people and machines are equivalent in all respects-they clearly are not. But when it
comes to making choices, so far, at least, there are not good reasons to believe they operate
according to different natural or scientific principles.
Either both people and computers can have free will, or neither can-at least until we discover
some evidence to the contrary.
31
ARTIFICIAL INTELLIGENCE Philosophy of Artificial Intelligence
be able to imitate human action, but they would lack human kind of consciousness. For a
machine that learns on its own, there needs to be a separate debate on whether it would
be conscious or not. Or would it be mathematically and algorithmically conscious? As a
reader of this book, you can make your judgement
The notion of consciousness or subjective experience more generally, simply does not apply
to machines.
The answers to our questions hinged largely on whether you regard words like intelligence,
thinking, and feeling as connoting something sacrosanct about humans (or at least biological
creatures), or whether you are comfortable expanding their applicability to certain artifacts,
gadgets or machines.
The challenge posed by AI is how to describe, and therefore how to understand and reason
about, a phenomenon that has never been encountered in human experience-computational
devices capable of perception, reasoning, and complex actions. But the words that seem to fit
these new developments most closely are colored with implications about humanity’s uniqueness.
It is relatively simple to build a robot that flinches, cries out, and / or simply says, “Ouch,
that hurts” when you pinch it. Since computers are so different from us (at least today) and
are designed by us for specific purposes (as opposed to naturally occurring), it seems logical
to say they do not, and most likely never will, have real feelings. Although within this book
there is an example of robot that mimics human emotions, it is grossly unfair to say (as of
now) that computers will end up resembling human beings in aspects of feelings and emotions.
Hence, to conclude we may say that Artificial Intelligence, as any other machine we may
know of in the world can make us lives easier in terms of what purpose we have with the
technology. Of course in many cases an AI machine would understand our moods, emotions
or even feelings when we use the technology for say enhancing mental well-being. Or in
the Military domain, machine learning could perhaps predict human behavior based on
demography, standard of living etcetera while looking at potential terrorist threat. AI has a
use wherever we want to use if all the ethical boundaries by nations are maintained.
Activity
32
ARTIFICIAL INTELLIGENCE AI and Future jobs
The actual process by which machines displace human workers is much more subtle. In
practice, automation replaces skills, not jobs, and correspondingly, what employers need
is not workers, but the results obtained by applying those skills. To be successful, makers
of robots do not have to replace people; they have to provide machines with the requisite
skills to perform useful tasks. And while their products may not replace workers one to
one, they nonetheless put people out of work because fewer workers are required. But the
process also changes the jobs of those still employed by eliminating the need for a particular
skill, and possibly adding requirements for new skills.
A good example of this process is as close as the checkout stand at your local supermarket.
The clerks who total your bill (cashiers) and package your groceries (baggers) are engaged
in several skill-based tasks that have changed significantly over the past few decades. The
cashiers used to examine every item in your shopping cart and key the price into their cash
register, whereas now they simply swipe the items across a bar-code reader. The advantages
of the new system in terms of accuracy, time, and labor are obvious. But the main reason
the cashiers are still there to greet you is that certain items require special handling. In
particular, bags of loose produce need to be identified and weighed to determine a price.
And these are skills that so far, at least, have resisted automation.
Nothing about AI changes the fundamentals of how labor markets evolve with technology.
From an economic standpoint, AI technology is just another advance in automation. But its
potential to rapidly encroach on current workers’ skills is unparalleled in the recent history
of technological innovation, except for the invention of the computer itself.
33
ARTIFICIAL INTELLIGENCE AI and Future jobs
unique skills a given worker utilises, the more vulnerable he or she may be to replacement
by a machine, depending on the skills, of course. But even if only a portion of a worker’s
expertise or experience is amenable to replacement, improving productivity has the effect
of reducing overall employment.
So yes, robots are going to take our jobs, but a more useful way to think about it is that
they are obsoleting our skills, a process economists call “de-skilling,” appropriately enough.
And there is nothing new about this process-the magnitude of the impact of AI in particular
will depend on how quickly and widely the new technologies will facilitate automation
of workers’ skills. And on that front, the news is neither good nor bad. If you are in a
technologically superior nation like the United States or countries in Western Europe, the
chances of many jobs becoming obsolete is very High compared to the jobs in Countries
like Cambodia, Pakistan, or any of the technologically inferior nations
There is another broad class of jobs in which we employ people just to pay attention. The
potential of facial recognition in crowds to flag possible suspects in well known, but the
accuracy and broad deployment of such systems is increasing dramatically, heightening privacy
concerns. Visual recognition systems will be able to identify and discriminate permitted
from prohibited activities, such as a customer in a store walking into an area restricted
to employees or attempting to carry out merchandise without paying. Several supervisory
functions fall into this class. For example, an AI system will be able to summon a waiter to
refill a customer’s water glass or clear a table for the next guest. One computer vision system
currently being tested on the Stanford University campus counts the number of people
entering a restroom in order to customise the schedule for attendants to service the facility.
Future traffic lights will be able to anticipate your arrival, change dynamically to facilitate
vehicle flow, and stop traffic when a pedestrian or obstruction (such as a dog) appears.
34
ARTIFICIAL INTELLIGENCE AI and Future jobs
Historically, the jobs most susceptible to automation have been characterised as those that
are routine or those which involve performing the same activity or task repeatedly or, with
the advent of computers, are readily described in an explicit set of steps or rules, and so can
be more easily reduced to a programmatic formulation. But AI is expanding this constraint
to many tasks that are decidedly less routine. For example, successfully driving a car may be
well defined, but the task is hardly routine. The same could be said for reading handwritten
documents or translating text between languages. Yet, machine learning techniques have
proven very adept at these challenges, often equaling, or exceeding human capabilities.
Using so-called big data, many tasks that might otherwise appear to require insight and
experience are now within the purview of today’ s or tomorrow’s machines. Indeed, detecting
patterns too subtle or fleeting for human analysis, such as the flow of data in a network,
the movement of potentially hostile troops near a contested national border, or the activity
of credit card charges signaling the possibility of fraud, is now practical.
Industrial robots have long been used for simple, repetitive tasks like welding and assembling,
but recent breakthroughs in sensory systems permit these mechanical servants to escape the
factory floor and seek wider employment. The missing part is the “brains.” We can connect
low-cost sensors up to dexterous manipulators but translating the stream of data into action
is another matter. This is a hard problem, but AI engineers have a trick up their sleeves:
plenty of economically valuable undertakings can be deconstructed into a series of smaller,
simpler tasks, each performed by separate devices and techniques. Complex tasks such as
installing a new lawn sprinkler system can be broken down into more easily automated
components. One robot may deliver the materials, another may dig the ditch, a third might
lay and connect the pipe, and a fourth might backfill. A human worker might still be needed
35
ARTIFICIAL INTELLIGENCE AI and Future jobs
to design the layout and supervise the work but that’s cold comfort to the myriad manual
laborers previously required to complete this job. Industrial and commercial automation
on a grand scale does not require some magical recreation of human intelligence- it just
has to get the job done.
And that’s where AI comes in. For the most part, one-off solutions that do not generalise
are perfectly fine as long as they deliver economic value.
At the current state of affairs, there is no fundamental scientific breakthrough required for
an Al-based system to tackle the vast majority of blue-collar jobs as the constraint is mostly
just painstaking engineering. If the task is well defined, relies on available sensory input,
and falls within the operational capability of available mechanical technology, it’s simply a
matter of time until some clever inventor configures the components and writes the software
to replace human workers.
• sewer diggers
• watch repairers
• machine operators (many different subcategories)
• tellers
• shipping, receiving, and traffic clerks
• drivers
• inspectors, testers, sorters, and samplers
• projectionists
• cashiers
• grinders and polishers
• farm laborers
• lobby attendants, ticket takers
• cooks
• gaming dealers
• locomotive engineers
• counter attendants (at cafeterias, coffee shops, and the like)
• postal clerks
• landscapers and groundskeepers
• electrical and electronic equipment assemblers
• print binding and finishing workers
36
ARTIFICIAL INTELLIGENCE AI and Future jobs
While it may be possible to automate these professions in the near future, it’s worth noting
that it may not always be desirable to do so as in many cases, the value delivered by a
worker is social as well as physical.
By contrast, there is a similar selection of blue-collar jobs that the Oxford study lists among
the least susceptible to automation:
• recreational therapists
• audiologists
• occupational therapists
• orthotists and prosthetists
• choreographers
• physicians and surgeons
• dentists and orthodontists
• fabric and apparel patternmakers
• athletic trainers
• foresters
• registered nurses
• makeup artists
• pharmacists
• coaches and scouts
• physical therapists
• photographers
• chiropractors
• veterinarians
• fine artists and craft artists
• floral designers
White-collar jobs are characterized by the processing of information, so many of them are
natural targets for automation by a computer. Some are engaged in a rote process, such as
transcription of handwritten notes into electronic form. Sometimes this process requires
skills that come naturally to humans but are more difficult for machines (at least today),
such as converting spoken language into written words. A subset of white-collar workers
is so called knowledge workers, whose main value is expertise but whose output is still
information, such as software engineers and accountants.
37
ARTIFICIAL INTELLIGENCE AI and Future jobs
In some ways, applying AI techniques to white-collar tasks is less challenging than to blue-
collar tasks. As a general matter, manipulating information is easier than integrating with
the physical world, and more natural for computers. In addition, white-collar tasks tend
not to be as real-time as blue-collar tasks.
The Oxford study includes the following selection of white-collar professions as among the
most susceptible to automation:
• taxation consultants
• title examiners
• insurance underwriters and claims processors
• data entry and brokerage clerks
• loan officers
• credit analysts
• bookkeeping, accounting, and auditing clerks
• payroll clerks
• file clerks
• switchboard operators
• benefits managers
• library assistants
• nuclear power reactor operators
• budget analysts
• technical writers
• medical transcriptionists
• cartographers
• proofreaders
• word processors and typists
And the study counts the following among those the hardest to automate:
• computer systems analysts
• engineers
• multimedia artists and animators
• computer and information research scientists
38
ARTIFICIAL INTELLIGENCE AI and Future jobs
• chief executives
• composers
• fashion designers
• photographers
• database administrators
• purchasing managers
• lawyers
• writers and authors
• software developers
• mathematicians
• editors
• graphic designers
• air traffic controllers
• sound engineers
• desktop publishers
Missing from the above lists are what are called “pink-collar” workers. These are people who
work primarily in service industries where face-to-face contact is an essential component of
their duties, or in which the observation or expression of human emotions is important.
Examples are waiters and waitresses (who provide table service as opposed to simply processing
orders), clinical psychologists, police, administrative assistants, classroom teachers, real estate
agents, consultative sales professionals, clergy, supervisors, and nurses. While some aspects
of their jobs may be subject to computerization, the remaining portions mainly those that
require an intuitive connection with other people-are likely to resist being automated for
the foreseeable future.
Activity
References
1. https://www.stottlerhenke.com/artificial-intelligence/history/
2. https://www.oxfordmartin.ox.ac.uk/publications/the-future-of-employment/
3. https://link.springer.com/book/10.1007/978-981-15-6548-9
4. https://www.investopedia.com/terms/a/artificial-intelligence-ai.asp
5. http://jmc.stanford.edu/articles/aiphil2.html
6. https://ai.google/education/
7. https://www.analyticsinsight.net/top-10-artificial-intelligence-innovation-trends-
to-watch-out-for-in-2021/
8. https://appen.com/blog/the-latest-innovations-in-artificial-intelligence/
39