Electric brains

I have been reading a book by Gerald Edelman called 'Second Nature' (2006), in which he follows in the foot steps of many an eminent scientist in spreading the wings a bit and trying for a theory of everything (aka a toe) before passing on the baton.

A decent rather than good bit of book production from Yale, certainly not up to the usually much higher standard of Belknap from Harvard.

The first part of the book contains a handy summary of Edelman’s generally convincing thoughts on the neuronic basis of consciousness. It also makes quite a big issue on pages 21-22 of asserting that the brain is not a computer, an assertion mainly founded on brains being very much larger computers than any man made one.

First, one has the complexity of an individual neuron and its firing arrangements - a rather more complicated object than a transistor. And then one has a huge number of them and an even huger number of connections between them. But a quick peep at Wikipedia tells me that a modern chip might contain the equivalent of billions of transistors, so I am now unsure how the 30 billion or so neurons in the brain compares. I had not realised that chips got anything like that big.

Second, one has the distribution of processing in a brain. Every neuron is a law unto itself and they are all processing away in parallel, while a computer, roughly speaking anyway, processes exactly one instruction (from a program which has been written and fed into it) per unit time. One might argue that the effects of interest in a brain endure for hundreds of milliseconds and when we look at intervals of this sort a modern computer can be considered to be multi-threading – which is true but the multi-threading is maybe hundreds of threads – not the millions if not billions of the brain.

Third, one has the evolution of the individual brain. We might have a genetic start point, but what happens after that depends on the environment with every environment being unique –  and different environments will result in different choices being made and different wiring up. A pair of identical twins might start with the same genes, but by the time they are born their brains will be different; less different than those of other siblings, but different nonetheless and giving different answers. Whereas one computer is pretty much like another: within limits, if you run the same test on different computers you will get the same answers.

Another point here is that the evolution of the neurons of an individual’s brain during that individual’s growth and lifetime result in what amount to computer programs which would be very hard to design, top down, from scratch. A consideration which I believe is already taken into account in, for example, robots which walk; they learn to walk, they train their neural network, in rather the same way as a human.

Fourth, one has the indetirminacy of a brain. Again within limits, if you run the same test on a computer two times you will get the same answers. This is not true of a brain which is continually changing and which is hosted by, interacts with an environment which is itself continually changing; you cannot rely on getting the same answer two times running – although our language skills do make this a bit more likely than it might otherwise be.

OK, so a brain is not much like a computer. But so what?

We can certainly model the workings of brains on computers (something which Edelman’s own institute does a lot of. See http://www.nsi.edu/), although it remains to be seen how close to a brain a computer is going to get. There are certainly a lot of people out there doing their level best to get the computer to beat the brain and I for one believe that, for many purposes, the computer will win. Rather more doubtful is whether such capability will be put, at least most of the time, to benign use.

Rather more doubtful also is whether a model of a brain on a computer will ever exhibit consciousness in the way of a human. Will the different engineering mean that we are never going to get there? The robot might well come to make a nice companion but will it remain firmly on the machine side of the divide? Who knows, but that is not going to stop lots of people from trying. A toe for all seasons.

On the other hand, there is, I think, good news here for the Freudians. I think this book, and others like it, show that there is going to be both need and room for the psychoanalytic way of describing the goings on in the brain. Maybe without the structures psychoanalysts use now, for example the oedipus complex, but with their approach and their methods. Their peeling away of layers of mind to get at beginnings and the beginnings of problems. They will have another day in the sun!