You're right, there are two questions: why haven't people tried very
hard so far, and is it worth trying now.
>"We don't know how long it takes"? Perhaps, but I think believing an
>agressive species could be noticeably domesticated in less than 10
>generations is like John Clark's claim that godhood will be feasible
>seconds after the first nano-assembler. And 10 generations for an ape
>is the better part of a century. And I suspect it would take more
>generations... you need to be improving docility and immunity and
>utility simultaneously. You might be able to get a few docile animals
>quickly, but the gene pool will be small. If you try to get a pool of
>animals who don't get sick too easily outside of their natural
>environment, you're spending a lot of time on agressive, non-useful
>(low-return) animals. I don't buy it.
I started this whole thread by describing existing lab monkeys, who
are smart enough, less threatening than wild monkeys, and don't get
sick that easily. We are already part of the way there.
>Why should an AI be hard to "domesticate"? There are evolutionary
>reasons for lots of the factors behind an animal being too vicious or
>panicky or territorial; there'd be no reason for AIs to have these
>traits. I think a major predator of chimps is other chimps, as with
>humans; AIs won't have evolved any defensive or offensive strategies.
We don't know what sort of evolutionary environment will be successful
in producing AIs. I do hope you're right that a non-competitive
environment will be sufficient, but we don't know.
>AI has a long road, but seems able to move very quickly. At least
>once it finds the road.
What indications do we have that AI can move quickly? I admit its a
logical possibility, but beyond that what do we know?
Robin D. Hanson hanson@hss.caltech.edu http://hss.caltech.edu/~hanson/