Books about Brains and Machines

Phantoms in the Brain: Probing the Mysteries of the Human Mind, by V S Ramachandran and Sandra Blakeslee

This book has a bit of a split personality, both in terms of its structure and writing. How much this has to do with the fact that it has two authors I'm not entirely sure, but the effect was to undermine Ramachandran's credibility in my mind. This the first of his books I have read, and may not have been a good place to start. However, it was weak enough in some ways that it makes me less likely to want to read more.

The first half is pretty much what I was expecting: a set of interesting anecdotes about the experiences of the patients of Ramachandran and others, and how these have illuminated our understanding of the workings of the brain. We hear stories about amputees with phantom limbs (and how the latter have been removed), people who believe that their parents are imposters, patients who cannot see anything on the left side of the world. These stories are told compellingly (presumably Blakeslee's contribution), and the way they are connected to the science is generally convincing.

Of these, one of the more memorable examples was the woman who developed sexual feelings in her phantom foot. It turns out the parts of the brain that process sensory information for the feet are adjacent to those for the sexual organs (according to the Penfield map). Assuming that the latter brain area expands thanks to the sensory vacuum created by the missing leg, it seems reasonable to think that the feeling in the phantom limb might be a mixture between the two. This may also account for foot fetishes. Interesting.

The second half is much more speculative, with Ramachandran explaining his hunches about areas that, as yet, have little evidence to back up any particular theory. Having heard Ramachandran speak about the brain before, I was actually expecting to enjoy this section, but this did not turn out to be the case: mainly because I was simply not convinced by his arguments. An element of this, no doubt, was his inability to be clear about where he stood on the issue of God: the subject of one of his chapters. This I found slightly cowardly, as if he were trying to avoid alienating either the faithful or the athiest elements of his potential readership. I will come clean and say I fall into the latter camp.

More importantly, I was very disturbed by a section in which he first explained why scientists were rightly suspicious of thought experiments, and then proceeded to describe a truly terrible one. He suggests if you were color blind he might be able to get you to feel the sensation or qualia of red not by just knowing that light coming in has a particular wavelength but, by "hook[ing] a cable of neural pathways from the color-processing bits of my brain to the color-processing bits of your brain." He claims that, though far fetched, there is, "nothing intrinsically impossible," about this scenario.

But how would a color-blind person develop color-processing bits of the brain? Though we know the brain has an architecture that determines which bits are used for what general tasks, we also know that elements of the brain that are not used atrophy (like the sexual organs over-writing the bit of the brain previously allocated to the foot). For instance, those born blind do not have wasted visual cortices: this section of brain is devoted to the remaining senses. In fact, it could be argued that someone's inability to experience certain qualia is caused by the fact that they never developed the appropriate circuits in the brain to process that 'feeling'.

When reading sections that, to me, seemed so illogical, I wondered whether some dumbing down had occurred in order to make the book sufficiently 'popular' science. However, it feel it unfair to blame Blakeslee when, at best, Ramachandran went along with this strategy (if that's what it was). Don't misunderstand me: there was far more in it that I found sensible than didn't (and pretty much all of it was at least interesting). However, there were enough fairly major deviations from reason as I know it that it undermined my trust in Ramachandran's analysis.

Another weakness in the second half of the book was the explanation of Ramachandran's theory about the elements that make up consciousness. I won't go into detail about these elements because he didn't... and that to me was the problem. The definitions he gave were far too lightweight for my taste: the issue, for instance, of what 'the self' is in the context of an embodied creature was given just a paragraph of explanation, when the subject could easily fill a chapter of its own. Of course, that would be a different book, but a clearer elaboration might have convinced this reader in a way that the current version didn't.

For my blood pressure, I wish I'd read the just the first half of the book with his fascinating patient histories and experiments. The more speculative stuff might also have been interesting, but only if taken seriously and explained properly. If this couldn't be done for a popular science audience, then perhaps it should not have been included in a popular science book.

Being There: Putting Brain, Body, and the World Together Again, by Andy Clark

Of all the books I’ve ever read that related to artificial intelligence, robotics, cognitive science, neural networks, and the brain, I’m pretty sure this is my favorite. Usually, for me, these books are about filling in the blanks in terms of how other researchers see the world. I get frustrated because I feel that their assumptions are wrong, their view is too narrow, or they lack imagination. I need to know about their view, but it’s not one I share.

With Being There, on the other hand, I felt I’d finally found my intellectual home. I probably agreed with 90% of it, and felt that Clark and I were coming at the problem—how human intelligence works—from the same angle: that without considering the body and environment, studying the mind and brain won’t get you very far. At the time this was still a minority, emerging idea in opposition to the ‘top down’ approaches to intelligence that had dominated AI for decades. Clark (now at the University of Edinburgh) previously taught at COGS at the University of Sussex, a world center for this way of thinking about the problem, so he was ideally placed to tell the story: it’s now become mainstream.

Clark starts by giving us an introduction to ‘the new robotics’, describing what were then (1997) relatively new approaches to building intelligent machines. These were successful because they weren’t focussed on the idea of planning: where a robot would map the environment, detect obstacles, plan a route, and then start to move. This traditional AI approach was very slow and unreliable because by the time the machine had decided on it’s strategy and started to implement it, the world might well have changed. It would then have to start planning again. Instead, the approach pioneered by Rod Brooks (discussed in an earlier blog) showed the effect of concentrating on just letting the machine get on with it and seeing what happened: potholes, obstacles, etc. could be dealt with at the time. Can’t move forward: move backwards or sideways. Foot not supporting weight: put your weight on your other feet. In other words, use the environment to do your planning for you.

But don’t get the idea that this is a book about robotics. The bulk of the book is a description of how the behavior of most biological species is inextricably linked to their physicality and their environment. From babies learning to walk against the force of gravity (leg weight affects behavior), to slime molds coming together to follow temperature and chemical gradients towards better nutrients: he gives lots of great examples.

One of the ideas I found the most interesting when I read this the first time is how we organize our environment to make life easier for our minds. (The environment doesn’t just control us, we control it.) For instance, we put all the spices in the same place, which means we only have to remember the location of one thing, not several. We put a letter on the table on the way to the front door to remind us to post it. And if we get Alzheimers and lose our memories, we find we can still function well as long as we stay in our longstanding homes that act as external memories. He calls this scaffolding, and once the idea gets in your head you see it everywhere: in informational devices like books and computers, conveyor belts, ergonomic design. In the end, he even links this idea up with language.

The book is not really straight pop science: it’s a little more academic than that. But it’s a very good and fairly easy read nonetheless. I found it particularly striking because I had tried to read one of Clark’s earlier books with MIT Press: Associative Engines. There I found sentences the length of paragraphs and paragraphs that filled pages. I don’t know whether he found a great editor, took a writing class, or what, but there were no such problems here. For students and scholars, not only does he have citations (that can be ignored if that’s not your thing) but he also is really good about giving credit where credit's due within the main text. Reading this book is the equivalent of taking an engaging but easy starter course in this approach to cognitive science, neuroscience, philosophy, and machine intelligence. It’s rich and satisfying.

In a short review like this it’s not possible to do this book justice. All I can say is that those who find my blog interesting should go out of their way to read it.