Chapter 1 Sample

Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/338843576

Artificial Intelligence for Everyone

Book · January 2020

CITATIONS READS

2 8,734

1 author:

Steven Finlay
Lancaster University
68 PUBLICATIONS 818 CITATIONS

SEE PROFILE

All content following this page was uploaded by Steven Finlay on 22 May 2020.

The user has requested enhancement of the downloaded file.


Relativistic
e-mail: AI@relativistic.co.uk

Copyright © 2020 Steven Finlay.


All rights reserved.

First published in Great Britain by Relativistic.

Unless permitted under UK copyright law, this work may only be


reproduced, in any written, printed or electronic form, with prior
permission of the publisher.

ISBN-13: 978-1-9993253-1-2

ISBN-10: 1-9993253-1-1

No plants or animals were mistreated in the writing of this book.


To Sam and Ruby

iii
Contents
Acknowledgements v
Foreword vi
1 Setting the Scene 1
2. What is Artificial Intelligence? 9
3. Algorithms, Models and more Algorithms 32
4. The AI Explosion. Neural Networks and Deep Learning 56
5. A Bit Further down the Rabbit Hole… 80
6. AI at Work 93
7. AI in Society 115
8. Ethics and the Law 130
9. What Does the Future Hold? 156
Appendix A. Further Reading 165
Appendix B. Glossary of Artificial Intelligence Terminology 168
About the Author 187
Notes 189

iv
Acknowledgements

I would like to thank my family, and my wife Samantha in particular,


for their support in writing this book.

v
Foreword

Artificial Intelligence (AI) is an important subject that I think


everyone should know something about. Not least, because it’s
having an impact on everyone’s lives every day – even if most people
don’t realize it.
However, most of the books I have come across seem to fall
into one of two categories. The first are technical books, targeted at
people with mathematics and IT degrees, who want to build artificial
intelligence applications. The second type of book is aimed at the
general reader, but to me, they seem to nearly always be overly
evangelical in their approach. There is a lot of overhyping, too much
emphasis on what might happen in the future and getting carried away
by the propaganda put out by some of the tech giants of Silicon
Valley.
There seems to be very few books that have tried to adopt a
more realistic and pragmatic approach. In particular, books that are
focused on the here and now, that seek to explain what Artificial
Intelligence is, how it works and how it is being applied in the world
today.

Hopefully, this is the book that will fill that gap.

Steven Finlay, January 2020.

vi
1. Setting the Scene

“Americans worship technology. It's an inherent trait in the national


zeitgeist.”1

“Self-Learning Machines,” “Algorithms,” “Deep Neural Networks,”


“Robotics,” “Automation.” These are just a few of the terms being
used to describe the seemingly endless torrent of new “Intelligent”
tools, apps and gadgets that are sweeping across the world at an ever-
increasing rate.
By all accounts, the big driver of these new technologies,
“Artificial Intelligence,” or “AI” as it is commonly abbreviated to, is
already influencing or changing almost every aspect of our lives. This
spans everything from how we work, travel and shop, to the way we
obtain news and information, to the gadgets in our homes. It’s also
having a very significant influence on the human relationships we
have with each other, how we communicate and how we express
ourselves.
According to some, AI is awesome! Give it just another couple
of years and no one will own or drive their own car. Google, Lyft or
Uber’s self-driving taxis will take us wherever we want to go. We
won’t need an army anymore because our borders will be secured by
autonomous robot fighting machines. Medical apps, with a perfect
bedside manner, will diagnose and even pre-empt our every medical
need, delivering individually tailored treatments based on our DNA
profiles and what our smart watches are reporting. Meanwhile, we’ll

1
Artificial Intelligence for Everyone

all be living a life of leisure while a plethora of robots and digital


assistants manage everything on our behalf. If you do happen to
want a job, then a robot recruiter will interview you and make an
impartial decision, with no preconceptions or unconscious biases
about your age, gender, race or religious beliefs, as to whether you
are right for the role or not. Utopia will have arrived!
But how much of this ultra-optimistic vision of a perfect
machine dominated world will actually come to pass? Where and
how quickly? What are the risks and dangers? Will it be a utopia or
dystopia? Will we control the machines or will they control us? Does
technology make us all more equal or exaggerate social inequality and
isolation? What ethical questions do these technologies raise and
how should society act to curtail their use? Is technology inherently
racist and/or sexist? Could a robot become the leader of the Ku
Klux Klan or another Hitler? What safeguards are required to stop
this happening?
These are just a few of the questions that can arise when
discussing these new AI-based technologies.
I have an opinion on all these questions, but unfortunately, I
can’t answer them for you definitively – no one can. Predicting the
future is a tortuously difficult task. Even the best futurologists
frequently get it wrong, and we should all be very cautious about
believing anyone who claims they know precisely what will happen.
New technologies are notoriously overhyped initially, even if they
end up being really useful eventually2.
A great example of this overhyping can be seen in the history
of self-driving cars. For years, the tech companies were telling us that
these things were just about ready to go. Every year, promises were
made about self-driving cars being available by the end of the year,
or early the following year, but each year the dates were pushed
back3.
What it looks like today, is that rather than the promise of
owning a truly self-driving car that won’t require a driving license to
use, the near-term reality is self-driving taxis. These operate in well-
defined “Geo-fenced” areas of a few major cities, overseen by

2
Setting the Scene

human operators at taxi central. A self-driving car today doesn’t


mean owning something that sits in your garage at home that can
take your children to school without you or whisk you across states
on vacation. Sure, we are going to get there one day, but it’s not a
universal reality at the moment and is unlikely to be for quite a few
years yet.
OK. So, it’s impossible to predict the future with certainty.
Also, it’s important to remember that, unless you are of a fatalist
persuasion, the future isn’t fixed – it can be changed by our actions.
We can all hold an opinion and draw our own conclusions. We can
then campaign for or against what we believe, and for those who live
in a democracy, vote for people who hold similar views to our own.
It has been said that all forms of government are ultimately flawed,
but most will listen to public opinion and will try to act in the interest
of its citizens when new situations present themselves. We can see
this principle in action with the backlash that has been seen against
facial recognition software and the action that many governments
and other jurisdictions are taking to ban or limit its use.
However, to be able to debate a topic in an educated and
informed way, and hence come to a reasonable conclusion about it,
you need an understanding of the thing being debated. AI is no
different. To discuss it in a sensible way you need to understand the
different types of artificial intelligence that exist and how AI works
at a basic level (you don’t need to be an expert). It makes sense to
know what it is capable of and what its limitations are. What is real
in the here and now? What’s just around the corner? And what is
pure speculation, about some distant future that your great-great-
great-grandchildren might experience?
If your view of artificial intelligence is as some sort of magical
techno-mysticism that does lots of clever stuff powered by teams of
miniature unicorns, then who knows what nonsense you’ll end up
thinking. If I’m going to debate what the speed limit for cars should
be in urban areas, then it makes sense that I know what a car is. I
need to know what benefits cars deliver, roughly how fast they can
go, the damage a car can do if it hits someone, plus some reliable

3
Artificial Intelligence for Everyone

information about the scale of the problem; such as the number of


car accidents. Otherwise, hey, I could end up arguing that as long as
the limit is no more than 2,500 miles an hour then no one is going
to get hurt, right? What’s your opinion on that? Do you agree?
Perhaps the most important thing to appreciate, and one of the
biggest mistakes made by many tech workers and some futurologists,
is to forget that technology does not operate in isolation. Society,
morality, law, religion, business practices, human behaviour and our
fears and prejudices all have a part to play. Technically, there’s no
reason why I shouldn’t have my own private nuclear reactor installed
in the basement to power my home. In reality, there are many
practical reasons why I shouldn’t have a nuclear power plant under
my house and I don’t think I need to go into the precise details as to
why – it should be pretty obvious.
One of the key reasons why we don’t have more driverless
vehicles on the roads is that in the early days of their development
not enough consideration was given to all the “Soft issues” that
needed to be addressed. “If only we didn’t have pedestrians and
human drivers to worry about!” But, if anyone had given it any
thought, then they should have realized, right at the outset, what the
problems were. Banning people from crossing roads - crazy. The
idea of Americans be forced to give up their right to drive? That
would make removing the right to bear arms from the American
Constitution look like a picnic. The developers of self-driving cars
are now up to speed with these issues and accept that they have to
deal with human habits, customs and behaviours, but it took a while
for some of them to get there.
It’s easy for someone coming from the tech sector to blame the
politicians, and social and legal constraints as the reasons why their
technologies don’t become more widely adopted more quickly. Or
sometimes, maybe they think it’s all down to the irrational fears of
the plebs at large, who should just put their fears aside and get on
with things. But the principles that apply are no different to any other
type of invention. AI tech doesn’t have an opt-out to the normal
rules of engagement. If I create a super-duper new type of smart

4
Setting the Scene

phone, then I need to make sure all of the electrical components


conform to the relevant health and safety legislation so that I don’t
fry consumers’ brains. If I develop a great tasting new food additive,
then let’s get that fully tested before I start putting it into baby food
and so on. I don’t go about moaning that the law is an ass when it
comes to these areas. I make sure I understand what the laws are and
how they will affect me before rushing ahead with mass production
of my product. In fact, I should be talking about these things in the
very first design meeting, almost before I do anything else.
Delivering a commercial product to the market is very much the last
link in the chain.
The tech sector is learning but we can still see evidence of this
narrow “Tech first” focused thinking in practice. For example, with
the “Move fast and break things” philosophy that is popular with
some Silicon Valley entrepreneurs4. Basically, get the technology
stuff working and out to market as quickly as you can, and then
worry about any problems later. Take crypto-currencies such as
Bitcoin, and more recently, the Facebook backed Libra. Clearly, the
inventors of these crypto-currencies didn’t give enough
consideration to the role and function of the regulatory authorities5.
Consequently, there is considerable doubt as to how successful they
will be in the future. The barriers to their wider use have nothing to
do with the underlying “Blockchain” technology that Libra and
Bitcoin are based upon. Rather, the biggest challenge is satisfying
central banks and other authorities that they will not undermine the
stability of the world’s financial systems. Financial regulators have
huge powers. They will only allow a new currency to become legal
tender once all risks have been understood and mitigated against.
Without an appreciation of the context in which a technology
is being applied it’s also easy to fall into the trap of thinking that just
because it’s technologically possible it will happen. The fact that
someone can build an intelligent fridge, that knows when you are
running out of milk and reorders it for you automatically, does not
mean that customers actually want that feature in a fridge. Or maybe
they do, but not because they’ll ever use it to buy milk, but because

5
Artificial Intelligence for Everyone

they can tell all their friends and neighbors about it. It’s a status
symbol rather than something that’s practically useful. Goodness
knows the number of drivers I’ve met who bought cars with self-
park features who, after showing them off once, never used them
again, or the mountains of virtual reality headsets that now lie idle.
As you can probably gather, my view is that the technology side
of things is often the easy part. People and society are far more
complex and difficult to deal with. Don’t get me wrong, I’m not
saying artificial intelligence tech isn’t enormously complex, and it
only exists because of some very clever people, but the fundamental
elements of artificial intelligence systems are a lot easier to get to
grips with than the nuances of the human condition.
With that in mind, the goal of this book is not to preach about
AI and all the wonderful things that it might lead to. Nor is it my view
as to what will happen in the future. Instead, the aim is to act as a
guide to explain artificial intelligence and to put forward some of the
arguments about its use and misuse. By understanding what AI is
and how it works, this will help you form your own opinions as to
how to approach artificial intelligence-based technologies. It’s then
up to you to decide what the pros and cons are and what, if anything,
we should do about it.
With that in mind, the key things we are going to cover in the
chapters that follow are:

• Discuss what artificial intelligence is and how it works.

• Understand where you are already likely to encounter AI in


everyday life today.

• Describe how artificial intelligence can be used to the benefit


of individuals, business and society.

• Describe the risks and dangers, legal and ethical issues around
the use and misuse of artificial intelligence.

6
Setting the Scene

• Mull over some of the things that may (or may not) be on the
horizon in the next few years.

As someone who has become a little hacked-off with a lot of books


that seem to be to written by people who spend their time
worshipping at the techno-altars of Silicon Valley, I think I should
also be very clear as to what this book is not.

• It’s not a book for Geeks or Nerds (there a difference!6) No


technical stuff here my friend. All explanations are simple,
with no fancy Greek letters that those guys call math. There
is a very tiny bit of arithmetic in the book somewhere, but it’s
very much of the 2+2 = 4 kind. Not the E=MC2 variety.

• It’s not a pro-AI propaganda piece. Everything is not 100%


awesome in the world of AI. Sure, it’s really interesting, is
changing society in all sorts of ways and has some great
applications, but it’s not a universal panacea for all the world’s
ills and may even create some new ones.

• It’s not an anti-AI piece either. I’ve seen what the stone age
was like and I don’t want to go there. I want to move forward
not backwards. Something doesn’t have to be perfect to be of
benefit, it just needs to be better than what came before.
Where, on balance, AI can make life better, let’s get on with
it.

• It’s not written by someone employed by a tech giant or other


interested parties. The author, has worked in AI related fields
in the past, but regards himself as pretty impartial these days.
He can also confirm that he has not been paid by the Russians
or North Koreans to corrupt your thinking about American
technology companies.

7
Artificial Intelligence for Everyone

In terms of book structure, then it’s pretty much a game of two


halves. The next four chapters are very much about understanding
AI, what it is and how it works. The final four chapters are more
about how AI impacts what we do and the legal and ethical issues
that it raises.
Oh, and before I forget, if you see any text in bold italics, then
that means it has an entry in the glossary at the back of the book.
Fantastic. Now that we’ve got that sorted, let’s move on to the
next chapter and chew over what artificial intelligence actually is.

8
2. What is Artificial Intelligence?

“Any artificial intelligence smart enough to pass a Turing test is smart enough to know
to fail it”7

OK. So, I hear this term “Artificial Intelligence” being talked about
all the time, but what is it? It’s supposed to be everywhere but I
haven’t yet met an android that I can have a meaningful
conversation with. I keep hoping to meet one in a bar or at my next
book club meeting, but no luck yet.
There are loads of different and varied explanations as to what
artificial intelligence is but, I’m afraid to say, there isn’t one universal
definition that everyone agrees with. Not to worry. Let’s have a bit
of a poke around and see what we can come up with.
A simple and very practical explanation is that something
human-made, a machine, that can act and reason just like a person
embodies all that it means to be artificially intelligent. This isn’t an
unreasonable place to start and leads one to envisage a world filled
with human-like robots doing human-like things. Possibly our
servants, conceivably our masters, maybe our friends. But what
about dogs and dolphins? Don’t they display a degree of intelligence
also? Mmm. Let’s leave animal intelligence out of the debate for
now.
If we were to go back in time a few years, then this description
of human-like entities behaving in a similar way to ourselves is one
that many people would have readily agreed with without much
further thought. In fact, the ability to replicate human behaviour is a

9
Artificial Intelligence for Everyone

key feature of the famous “Turing test.” This was devised by the
mathematician Alan Turing,8 back in the mid-20th century, to
determine if a computer can be considered intelligent.
In the test, Turing envisaged a human in one room and a
machine in another. A human judge then converses with the human
and the machine remotely, without knowing which is which. The
judge may engage them in conversations about the weather, politics,
the latest clip they’d seen on YouTube or anything else they liked.
The machine is deemed to be intelligent if, after a period of time
spent in conversation, the Judge is unable to distinguish between the
human and the machine.
At the time of writing, no machine has passed the Turing test,
although there have been some very good efforts, and it will
probably be many years before a computer is able to do so
convincingly and repeatedly9.
The Turing test is pretty cool, but it’s not without its problems.
Lots of people have tried to write computer programs that mimic
human responses in an attempt to fool the judge and pass the test,
and when the test is finally passed, this is probably the route that will
be taken. The focus has been on passing the test, rather than creating
something that is actually intelligent; i.e. clever programming tricks
designed to deceive are not in the spirit in which the test was
originally conceived. To put it another way, these people who are
trying to pass the test haven’t built a really kick-ass robot, and then
said: “Oh, by-the-way, I wonder if it would be any good at the Turing
test? What do you think Robot, do you want to give it a go?” Rather,
they’ve said “Mmm, let’s see, what do we need to do to pass the
Turing test? and let’s try and build something that does that.”
Another argument against using the Turing test to assess
intelligence is that a machine doesn’t need to be conscious in the way
people are to pass the test. It doesn’t need to be self-aware of its
actions or have any understanding of the responses it makes. It can
pass by just blindly following some (very complex) computer code.
If we stick to the concept of a fully conscious self-aware
thinking machine, with all the different mental capabilities that we

10
Glossary of Artificial Intelligence Terminology

have, as our definition of artificial intelligence, then we are years, if


not decades, away from such machines. In fact, there is an argument
(that many scientists dispute) that no system, based on the types of
computers we use today, could ever be truly conscious as we
understand it, no matter how powerful they are. There must be some
additional (as yet undiscovered) element required for human-like
consciousness and self-awareness that can’t be replicated via
computation alone10.
Even if we remove the requirement to be conscious and self-
aware, and instead focus on the ability to mimic the full spectrum of
human intellectual capabilities using massively powerful computers,
then there is nothing anything like this in the world today. This broad
“Do it all” type of artificial intelligence, or General AI as it is often
referred to, is also likely to remain in the realms of science fiction for
some time to come.
Taking a slightly more philosophical line, isn’t it a little arrogant
to define intelligence as a solely human property? Could there not be
other, completely non-human (alien) ways of thinking that can also
be classified as intelligent? A machine that thought in this “other
way” would therefore be intelligent, but possibly not in a way that
we could readily comprehend.
So, fully human-like AI is still a distant dream. But don’t be too
disappointed. Instead, let’s lower the bar a little. Instead of thinking
of AI as something that must be like a person (or an alien), let’s adopt
a less stringent, and somewhat narrower, definition of artificial
intelligence. If a machine can reason, such that it can make decisions
that are at least comparable to those that a human would make, then
that’s artificially intelligent behavior. To put it another way, a
machine that can gather information and learn about a situation, and
then decide upon a sensible course of action based on what it has
inferred, is behaving intelligently.
If we want to split a few hairs, then saying that something is
“Behaving intelligently” may not be quite the same thing as “Being
intelligent” but let’s assume that that’s good enough for now. More
concisely, we can summarize this somewhat restricted definition of

11
Artificial Intelligence for Everyone

artificial intelligence as follows:

Artificial Intelligence (AI) is the ability of a machine to


assess a situation and then make an informed decision in
pursuit of some aim or objective.

This definition encompasses the replication of both conscious and


unconscious decision-making in humans and other animals. An
example of conscious human decision-making is where someone is
assessing the experience of different job candidates with the aim of
deciding who is best suited to fill a vacant role. A machine that could
do this instead of a hiring manager would qualify as being intelligent
within this area of expertise. An example of unconscious decision-
making is taking in the features of a picture before deciding that you
are looking at a cat rather than a cake. Again, a machine that can do
this would fall within our definition of artificial intelligence. Both of
these are examples of artificial intelligence that are in widespread use
today.
In displaying intelligent behavior, a machine can assess the
information it has to hand and make a decision on the basis of what
it knows – an informed decision. If things change; i.e. new and different
information is made available, then the machine will reassess the
situation and a different decision may result. For example, introduce
pictures of potatoes, and the object recognition system can learn to
identify not just cats and cakes but potatoes as well.
If we look at the landscape today, then there are lots of clever
systems that meet this definition of artificial intelligence. When you
here about some new application of artificial intelligence, this is
pretty much the definition that is being referred to. Now, I’m not
saying that the definition of AI that I’ve just presented is one that all
the experts will agree with or is theoretically complete. In fact, it
would probably have many of the original pioneers turning in their
graves. However, from a practical perspective, it pretty much covers

12
Glossary of Artificial Intelligence Terminology

all of the tools, apps and gadgets in the world today that the tech
companies describe as incorporating artificial intelligence.

The full version of this book is available in print and electronic


format at Amazon, iBooks and all good book stores.

13
Notes

1 Max Brooks. (2006). World War Z: An Oral History of the Zombie War.
Crown.
2 This is a version of Amara's Law, formulated by the American futurologist

Roy Charles Amara (1925 – 2007). This stated that: “We tend to
overestimate the effect of a technology in the short run and underestimate
the effect in the long run.”
3 For example, Elon Musk was reported to have been talking about

autonomous cross-country trips being available in 2017. General Motors


promised us that self-driving rides would be available in 2019 and Ford was
talking about 2021, which now looks unlikely. Matt McFarland. (2019).
“Self-driving cars: Hype-filled decade ends on sobering note.” CNN.
https://edition.cnn.com/2019/12/18/tech/self-driving-cars-
decade/index.html, accessed 02/01/2020.
4 This is part of a quote attributed to Mark Zuckerberg which goes: “Move

fast and break things. Unless you are breaking stuff, you are not moving
fast enough.”
5 Yves Mersch. (2019). “Money and private currencies - reflections on

Libra.” Bank of International Settlements.


https://www.bis.org/review/r190902a.htm, accessed 02/01/2020.
6 I didn’t used to think there was a difference between Nerds and Geeks

but there is a body of opinion that they are actually very different. I really
like the explanation by Laurie Vazquez on the Big Think website, which
goes like this: “A Geek is an enthusiast of a particular topic or field. Geeks
are “collection” oriented, gathering facts and mementos related to their
subject of interest. They are obsessed with the newest, coolest, trendiest
things that their subject has to offer. A Nerd is a studious intellectual,
although again of a particular topic or field. Nerds are “achievement”
oriented, and focus their efforts on acquiring knowledge and skill over trivia

14
Notes

and memorabilia.” https://bigthink.com/laurie-vazquez/are-you-a-geek-


or-a-nerd, accessed 02/01/2020.
7 Ian McDonald. (2004). River of Gods. Simon & Schuster.
8 Alan M. Turing. (1950). “Computing Machinery and Intelligence.” Mind

49, p. 433-460.
9 There is some debate as to what constitutes a pass and under what

conditions. For example, how the judge is selected, and how many and what
type of questions they can ask. In my view, the machine would need to be
able to pass the majority of the time, when questioned at length about a
wide range of topics by a person of at least average intelligence. A one-off
fluke win would not count in my book.
10 John R. Searle. (1980). “Minds, brains, and programs.” Behavioral and

Brain Sciences 3 (3), p. 417-457. Also see the arguments in the 2019 book
by Christof Koch: “The Feeling of Life Itself: Why consciousness is
widespread but can’t be computed.” And those by Roger Penrose in his
books: “The Emperor's New Mind: Concerning Computers, Minds, and
the Laws of Physics” and “Shadows of The Mind: A Search for the Missing
Science of Consciousness.”

15

View publication stats

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy