BOOK AND SOFTWARE REVIEWS


Boden, Margaret A. AI, its nature and future. Oxford: Oxford University Press, 2016. [8], 198 pp. ISBN 978-0-19-877798-4 £12.99/$18.95


Artificial intelligence has had a chequered history, from the high optimism of the Dartmouth summer research project in 1955, which was proposed on the basis that:

The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves. We think that a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer. (McCarthy, et al., 1955)

'Significant advance' has been a long time in coming and disappointment in the early years led to a reduction, and in some cases, the complete disappearance of funding for the field until, I think, the emergence of so-called expert systems in the 1980s. Sinnce then the field has had its ups and downs, fragmenting into a number of sub-fields whose participants don't event talk to one another.

Perhaps no one is better qualified that Margaret Boden to untangle the complexities of the current situation: she has worked for many years at the interfaces between philosophy and the study of the mind, and both of these and computing. With degrees in philosophy, psychology and medical sciences, she brings a wide-ranging intelligence to bear on the issues, rather than the narrow focus of the computer scientist

To accomplish her task in such a short book (the text occupies only 169 of the 198 pages) is quite astonishing, but it does lead to a series of quite dense arguments, and while this book is accessible to anyone wishing to learn about AI, he or she will have to keep their wits about them. This is not an easy read.

Boden's definition of AI is deceptively simple:

Artificial intelligence (AI) seeks to make computers do the sorts of things that minds can do. (p.1)

Ay, but there's the rub! The problem is knowing how the human mind does the sorts of things it does, in order to be able to replicate those things in a computer, or, more likely, a battery of computers. This is not to say that AI has failed: far from it. Boden points to a wide range of applications, many of which are embedded in everyday objects, that we do not notice. Others are highly specialised, like the systems employed by banks to monitor financial markets, once called expert systems, but more commonly known today as rule-based systems.

This short book has seven chapters: 1 - What is Artificial Intelligence; 2 - General intelligence as the Holy Grail; 3 - Language, creativity, emotion; 4 - Artificial neural networks; 5 - Robots and artificial life; 6 - But is it Intelligence, Really; and 7 - The Singularity. The totality, as you see, is very wide-ranging, suggestinge the diversity of problems with which the AI communities are wrestling, often from completely different perspectives.

The really crucial issues in AI are dealt with in Chapters 6 and 7: chapter 6 considers the nature of intelligence and whether the operations of a computer can be called intellligence, which gets us into deep philosophical and neurological issues. Does intelligence imply consciousness, and what is consciousness. There's no agreement on what consciousness is and how humans acquired it (whatever it is), so determining whether or not a computing machine reaches a conscious state is problematical. If we don't understand it in ourselves, how would we recognize it?

The singularity question is one that has seen some publicity recently in various warnings about the potential impact of AI on the human race (Sainato, 2015): the proposition is that at some point before the end of this century AGI or artificial general intelligence (which is not yet accomplished) will morph into ASI or artificial superhuman intelligence. At this point AI beomes either a threat to humanity, or it ushers in a brave new world with an absence of disease and, for some, the prospect of continuing life as our 'consciousness' is downloaded into computers.

Boden is sceptical about the prospect of ASI, as am I: one of the key propositions of the proponents of ASI is that because Moore's law demonstrates the pace at which computer processing power develops, so, ASI becomes inevitable. This seems to be a logical error of such magnitude that I don't understand how any serious scientist could hold it. It's rather akin to proposing that, because the world's population is continuing to increase, so the various land masses upon which we stand will inevitably increase to hold the multiplying millions. Before any advantage could be taken of the increase in computing power forecast by Moore's law, all of the problems that AI still has to solve, including the nature of human intelligence, human learning, the nature of human memory, the functions of every neurone in the human brain, etc., etc., would have to be solved.

u

This is a hugely stimulating and thoughtful book - and in such few pages. It deserves to be read by anyone wondering about the hype that often surrounds developments in AI and it should be on the reading list of every information science, cognitive science, and computer science programme.

Reference

Professor T.D. Wilson
Editor in Chief
June, 2016