Re: COMP: AI substrate

Eliezer S. Yudkowsky (
Mon, 26 Jul 1999 16:03:18 -0500 wrote:
> Eliezer S. Yudkowsky, <>, writes:
> > If intelligence was that simple, rats would be sentient.
> I question whether intelligence is so helpful that all mammals evolved
> to be as intelligent as possible. It seems to me that intelligence is
> of questionable value in the day to day life of an animal.

I think you're wrong. I seem to recall a chart, demonstrating that over the past hundred million years or so there's been a steady increase, not just in the size of brains, but in the ratio of brain mass to body mass.

> Consider the life of prehistoric humans. I don't see why someone
> marginally more intelligent than others will have much of an advantage in
> hunting or gathering. Most of what is needed is learned culturally, and
> those techniques have been optimized over generations to approximate local
> maxima. Sure, even ten or twenty thousand years someone domesticates
> a new animal or discovers fire, but these events are so rare they can't
> have any significant evolutionary input.

The evolution of human intelligence had nothing to do with tool use, IMHO. Human intelligence was the result of social competition for mates. First intelligence allowed people to wind up with the best mates. Then, since intelligence was an advantage, women and men started seeing intelligence as being sexually attractive - all the people who want a "sense of humor" in their mate, for example. This established a positive-feedback cycle; the more desirable intelligence was, the more successful the intelligent humans, and the greater the evolutionary advantage to find intelligence desirable. And because the competition was other humans, the race never stopped. This is the way peacocks wound up with huge tails that require insane biological investments, or deer bucks got unwieldy racks of antlers, and it's the way we wound up with brains that consume 20% of our ATP (body's internal fuel).

> My feeling is that people stumbled onto some cultural and physical
> changes which set up specific incentives for greater intelligence -
> bipedalism, the use of spears, fire, tools, etc. This brought us up to
> IQ 100. But that was smart enough for these purposes. IQs of 130 or
> 150 wouldn't make someone any better at using these tools.

Even so, if there was anything unnecessary about the intelligence we have now, any way to get it more easily, then we'd evolve smaller brains that did the same thing. 20% of ATP is a pretty penny. I stand by my statement that brains, wherever found on the evolutionary ladder, are *the* most highly optimized part of the biology - with the possible exception of immune systems.

           Eliezer S. Yudkowsky
Running on BeOS           Typing in Dvorak          Programming with Patterns
Voting for Libertarians   Heading for Singularity   There Is A Better Way