Artificial intelligence (AI) has proven more difficult to create than its original proponents expected. This admirably accessible book explains why, and considers what it will take to get truly intelligent machines.

This potted history isn't complete, as J. Storrs Hall, a fellow of the Institute for Molecular Manufacturing and author of several books on nanotechnology, documents in Beyond AI. In fact, says Hall, AI started nearly a decade earlier as cybernetics, and built on the ancient idea of feedback mechanisms. We don't call the field cybernetics now, thanks to personal fissures in the research group led by Norbert Weiner, who jointly coined the term. Cybernetics as they conceived it has, however, spawned research in many fields, from control systems to today's robotics.

Explaining all this background doesn't make the problem of creating artificial intelligence any easier. The problem, as Hall notes, is not that robots are machines; the problem is that we, unlike robots, are made up of trillions of molecular machines. In biology, he says, 'It's machines all the way down'. We have, as yet, no idea of how to build trillions of self-replicating, programmable chemical factories that can organise themselves into a unified whole.

As a fellow of IMM, Hall is, of course, at the extreme end of the spectrum when it comes to believing we can learn to build such things. Once he has reviewed the strands of where AI came from, he goes on to consider what it will take to get to truly intelligent machines. He also examines what some of the consequences might be, via subjects like search, designs for brains and so on. The end of the book contains some giddily optimistic scenarios: wealth and leisure for all.

AI is the kind of subject that technical books often render so complicated that no one outside the field can stand to read them. Reading Hall's book is more like boiling a frog by slowly heating the pot of cold water it's comfortable in. The clarity of the writing is extraordinary. One minute you're reading perfectly intelligible prose about Frankenstein, and the next you're reading equally intelligible prose about the question of whether tomorrow's robots could acquire a moral sense — but nothing in between ever seems difficult to understand. Writers, if no-one else, know how hard you have to work — and think — to achieve that kind of deceptive simplicity.

Copyright © 1995-2009 CBS Interactive Limited. All rights reserved
ZDNET is a registered service mark of CBS Interactive Limited. ZDNET Logo is a service mark of CBS Interactive Limited.