"...it seems to me that almost everything we think we understand about the brain is wrong. We know an enormous amount about it now and just about none of it makes the slightest bit of sense. That's a good sign, I think. It shows us we've been looking at the wrong page of the map.
Let me try to illustrate this with a thought experiment: Suppose I give you a very complex system to study – not a brain but something equally perplexing. You discover quite quickly that one part of the system is composed of an array of elements, of three types. These elements emit signals that vary rapidly in intensity, so you name these the alpha, beta and gamma elements, and set out eagerly to study them. Placing a sensor onto examples of each type you find that their actual signal patterns are distressingly random and unpredictable, but with effort you discover that there are statistical regularities in their behaviour: beta and gamma elements are slightly more active than alpha elements; when betas are active, gammas in the same region tend to be suppressed; if one element changes in activity, its neighbours tend to change soon after; gammas at the top of the array are more active than those at the bottom, and so on. Eventually you amass an awful lot of information about these elements, but still none of it makes sense. You're baffled.
So allow me to reveal that the system you've been studying is a television set, and the alpha, beta and gamma elements are the red, green and blue phosphor dots on the screen. Does the evidence start to fit together now? Skies are blue and tend to be at the top, while fields are green and tend to be at the bottom; objects tend to move coherently across the picture. If you know what the entire TV image represents at any one moment, you'll be able to make valid predictions about which elements are likely to light up next. By looking at the entire array of dots at once, in the context of a good system-level theory of what's actually happening, all those seemingly random signals suddenly make sense. "Aha!"
The single-electrode recordings of the equivalent elements in the brain have largely been replaced by system-wide recordings made by fMRI now, but at the moment we still don't know what any of it means because we have the wrong model in our heads. We need an "aha" moment akin to learning that the phosphor dots above belong to a TV set, upon which images of natural scenes are being projected. Once we know what the fundamental operating principles are, everything will start to make sense very quickly. Painstaking deduction won't reveal this to us; I think it will be the result of a lucky hunch. But the circumstances are in place for that inductive leap to happen soon, and I find that tremendously exciting."
Tuesday, March 06, 2007
Understanding the brain - an inductive leap?
This clip from a brief essay by Steve Grand, A.I. researcher: