Saturday, April 09, 2016

A Bunch of Dumb* Guys Arguing About IQ

Lubosh has joined the fray. And PZ Myers has another rejoinder to Steve.

Myers's latest is such a confused mishmash of dubious assertions, irrelevancies, and physicist baiting that it hurts me to say that he's mostly right, but he is. He claims, with no real evidence, that human IQ hasn't increased in the last 100,000 years. He makes a big deal of the fact that race horse speed hasn't increased in 50 or more years. This is true, but it ignores the fact that the effective breeding size of the race horse population is tiny - about 31 ancestors, and that artificial rules prohibit any kind of genetic engineering not known to the ancient Egyptians. He makes a big deal of the fact that super high IQs may not have been adaptive in the past, and that there are ethical and practical reasons why experimenting with genetic engineering of smarts might not to be a good idea.

Lumo makes the point that this latter fact is irrelevant - for the moment we are talking about possibility not practicality.

As I said, though, I'm pretty sure that Myers is mostly right, and moreover, that genetic tweaking of IQ is probably a dead end, except maybe for the purpose of preventing severe retardation. For one thing, we know little about what IQ or differences in IQ mean in terms of brain structure and organization. What would a person with an IQ 200 be able to do that lesser mortals can't, for example, aside from really crushing the Raven progressive matrices?

We know one key fact about the changes associated with human advances in cognition over the past several million years: size matters. Brain size tripled over this period, and brain size differences in current populations are correlated with some measures of IQ. So how did this work? Nobody really knows. On the other hand, size isn't everything. Neandertals had bigger brains than us but they were culturally more primitive and we exterminated them.

My guess is that the engineers at AlphaGo could whip out an AI system that would really, really, crush Raven's progressive matrices. Would that mean that that systems had an IQ a lot higher than humans? Technically, yes, but practically, hardly. The most obvious limitation of the human brain is its very slow cycling/processing speed, which is measured in milliseconds. Silicon based processors are a million times faster, and if we care about higher IQs, that's likely to be the way it could happen.

*Steve and Lubosh probably have IQs higher than mine, but that hasn't prevented them from being wrong about this.