NO BRAINER

Dick Pountain/04 May 1997/Idealog 33

I don't normally have either the time or inclination to participate much in online debates, but a recent discussion in Derek's Prolog conference on Cix caught my eye because it was about the brain, and that was my chosen subject back in the days when I still thought that I might end up as a biochemist. It was a well-mannered discussion - by no means what Mrs Merton might call a 'heated debate' - and it all started when a participant commented on Derek's May Prolog column which was about the superiority of human over computer memory (in deference to previous privacy squabbles I won't mention any names here, nor quote any Cix text directly.) This participant wondered how long it would be before we could expect to buy Terabyte cartridges that we could plug into our skulls to expand and augment our own memory.

Some folk reckoned that this would not be a good idea on moral grounds, because we would be unable to apply our critical and ethical faculties to information contained in such an artificial implant the way we can to ideas received through the normal channels. This is a powerful counter argument, if the technology were ever to be realised, but I couldn't resist stepping in to say that in my opinion it never would be realised, and this attracted the charge from a participant that I must be one of those people who think that there is something transcendental and God given-about the human mind/brain system. Actually my reasons for thinking that it can never happen don't involve any divine influences, unless you regard Information Theory as God, which is a least as plausible as most theologies.

My argument ran like this. I'm inclined to believe that the human brain is superior to any artificial device we are ever likely to make because it was 'designed' by evolution, and the amount of information (ie. negative entropy) that got crystallised into its structure, over millions of individuals for millions of years, would be fantastically difficulty to accumulate in any deliberate way. What we are getting really good at however is inventing devices that are much better than the brain for some highly specialised task (eg. floating point arithmetic, alphabetical sorting) and using these as peripherals to enhance our brains' capabilities through the wonderfully flexible 'bus' provided by our senses.

To me the two greatest wonders of the world are evolution by natural selection, and the human brain which is the most impressive product of it so far (apologies in advance to all you Dolphin fans). Evolution is the most powerful algorithm that could ever be devised, because it's the self-organising algorithm - if you can't solve a problem just evolve around it, expanding to fill every niche and crevice of whatever problem space you inhabit. When that space is the Real World itself, evolution has filled it with life forms that inhabit almost every cubic centimetre, from those red worms in the black depths of the ocean to the Deer Beetles that fly in the jet stream. But evolution by selection considered as an abstract process will work in any problem space, as programmers who've played with genetic algorithms well know, and it can produce equally astonishing results everywhere. Problems that are quite intractable to a normal algorithmic approach succumb when you attack them with millions of randomly mutating programs, and a method for deciding which one produces the better result (that's the selection part). Nowadays information theorists tend to use a spatial metaphor to think about evolution, involving hill-climbing in an abstract landscape where mountains represent the 'problems' to be solved, and selection is any test that can tell you when you are going uphill. In the biological world these mountains would represent ecological niches, like a desert or a smoking vent on the ocean floor, and going uphill means acquiring abilities that help survival in that niche. It's starting to look as though evolutionary processes in this abstract sense may be discovered at all levels of the universe, from the stars to the superstrings.

I don't believe the human brain has a divine origin, nor that it's necessarily unique - equivalent structures may have evolved on other planets, but such speculation doesn't interest me until they turn up in Camden Town (parochial, moi?) But I do believe that there may be some things that you can't 'manufacture' (literally 'make by hand', or by human labour) because it couldn't be done in any reasonable amount of time. Living things are self organising at a molecular level - they're not assembled by an outside agent, and they've taken millions of years to concentrate huge quantities of information that's not easily readable or usable by such an outside agent. To those Drexlerites who hope that nanotechnology will duplicate the process, I'd say given a million years, perhaps.

As for interfacing silicon-based memory components to the brain, how exactly would you go about that? Tap into the 'wiring loom' that is the spinal chord if you're feeling brave, but what do you then send down each nerve fibre? The truth is that most of the human brain is not like a Von Neumann computer at all, but much more like a real-time signal processing system. To pipe, say, some text from a RAM chip into the chord you'd have to turn the text back into sensory nerve impulses. The cheapest and most effective way to do that might be to turn the text into raster information, display it on an LCD screen and then let your eyes turn it into the appropriate impulses. Must try it sometime.