This was another of my ‘books of shame’ that I felt the need to re-read. I actually got quite a long way into it first time and I can’t recall why I put it down. The aim of the book is to explore the notion of artificial intelligence (AI), whether or not machines can truly “think”. In order to get to this question, Penrose first spends a lot of time (most of the book, in fact) looking at a wide variety of seemingly unrelated topics.
After an initial discussion of AI, Penrose launches straight into what is probably the hardest chapter to get your head round. It’s all about algorithms, Turing machines and the computability of mathematical problems. He doesn’t spare the detail with pages of binary digits and computer programming languages. It takes a long time to work through, but if you can brave it, there is much easier, and more enjoyable, science in later chapters.
Once you get over the initial hump, we ease back into some gentle maths with Penrose first outlining his neo-Platonic view of notions of reality (one I admit that I share with him). He does this via some very basic complex analysis, looking at the detail of the Mandelbrot set, though without going into too much depth for the casual reader. From here he looks at the world of classical physics and then quantum physics, giving the reader a general grounding in the basics of modern physics whilst every now and then alluding back to the premise of the book, essentially asking if a machine could ever be constructed that would be capable of making the intuitive leaps that humans have managed in coming to our present understanding of the cosmos.
For the most part, this should be readily understandable with a modicum of scientific education, though to someone who didn’t do maths or physics at A-level, much of it may be new and take significantly longer to get to grips with. But even the expert reader shouldn’t get complacent. Penrose’s approach takes much which we may be familiar with and turns it sideways, giving good reason to scratch our heads and think things through anew. The 2nd half of the chapter on quantum mechanics is, admittedly, a bit tougher to get through; the section on spin was where I found my bookmark from the first time I tried to read it and gave up.
After finishing with quantum mechanics he looks at the thermodynamics of the universe, a line of thinking which led, many years later, to Cycles of Time. He ponders over some ideas of quantum gravity but not to any depth that one might be satisfied with. For other takes on that, I’d recommend The Road To Reality (also by Penrose), Brian Greene’s The Elegant Universe or The Three Roads To Quantum Gravity by Lee Smolin.
Eventually, Penrose comes back to the question of AI. In order to do this though, he needs to look at the basic physiology of the brain. Now Penrose is a mathematician and a physicist; he’s not a neurologist. As such, this section of the book doesn’t come across anywhere nearly as strong as the rest of the book. It is clear that this is a written by an educated amateur in the field rather than an expert. For much more detail on how the brain works, I would recommend John Ratey’s A User’s Guide to the Brain.
One fascinating idea that Penrose puts forth is that what may distinguish human intelligence and consciousness is not our rationality, but our irrationality. If all people behaved in accordance with a strict rationality (though even most rationalists, myself included, exhibit some irrational behaviour from time to time) then the strong AI proponents might have more of a case. But the very evidence of irrational behaviour is what Penrose finds most interesting.
Ultimately, no firm propositions are put forward in this volume. The book ends with some musings and a tentative point of view. I intend to follow up, albeit not for a while, with Penrose’s later volume, Shadows of the Mind. In the meantime, what we have is a book which is very loosely about artificial intelligence, but which is really a book about the foundation of computing, along with a tour of some of the great ideas of maths and physics.