Saturday, April 6, 2013

Genius

I found a few interesting resources from this page that I wanted to highlight / bookmark. I am reading through Basic Neurochemistry (Siegel, 8th) right now and am interested as a scientist in the effort to model intelligence based on the brain. There are a lot of variables that people try to plug into equations to predict IQ: brain volume overall, correlations between the amount and/or ratio white matter (myelin sheathing), gray matter, folding, interhemispheric interaction, density, wave patterns...etc.

Now that all this money is supposed to be poured into the Brain Activity Map Project or whatever they'll call it, I just figured I'd put out my prediction:

We tend to think of the brain in computer-related terms. The speed at which it processes information. How its "circuits" are designed and how signals propagate through them. But there is something as a biochemist that leads me to reject the binary on/off nature. Rather than thinking of neurons as in an "on / off" state, and trying to model the brain like a CPU, it is important to realize that the strength of the signal is a function of chemistry. The actual electrochemical potential (voltage) that we are discussing here goes back to the concentration of ions and transmitters. And since this concentration can vary, so can signal strength.

So now imagine trying to model a very simple brain, like C. elegans, studied by Bargmann and others,  with only 300 or so neurons. In the simplest computing model, you would have the firing patterns of the neurons mapped so that, for instance, odor recognition lights up a certain portion of the brain in a certain sequence. But this assumes each of these neurons to be in a true "on/off" state. They are more along a spectrum of signal strength. Let's say for simplicities' sake you round off the observed concentrations into 10 brackets (zero, near-zero...near-maximum, maximum). The complexity inherent in mapping 300 different possibilities is exponentially magnified because it isn't 2^300, but instead it is now 10^300. So now you've gone from 2 * 10^90 to 1 * 10^300 possible "states". And if you aren't mathematically inclined, 10^90 is more than the estimated number of atoms in the universe. You would have to assemble 10^200 universes to approach the number atoms equal to the possible states in C. elegans brain.

Where would we put that kind of map, assuming a computer could draw it for us? ;) And do you think that reading it would be any different than an ant reading Melville?

As I've said before, I don't put too much stock in IQ tests that fall within one standard deviation or so from the mean, but I put a lot more in them when the measurement indicates significant outliers. I put zero stock in the effort to link brain activity meaningfully to behavior or intelligence.