From: Robin Hanson (hanson@hss.caltech.edu)
Date: Fri Sep 27 1996 - 22:13:44 MDT
Eugene Leitl writes:
>> Of course it is precedented. That is exactly what research feeding back
>> into education is. As humanity learns more, we get better at learning
>> more, and our understanding expands. It just doesn't explode as fast
>> as you think possible.
>
>You are right, of course. But the intelligence of the atomic agents has
>remained a virtual constant. Only the degree of cooperation has
>increased. But we're talking about autofeedback loop upon both the number
>of agents _and_ grain size of their intelligence. With a very short time
>constant, on a week-to-month scale.
>This is certainly unprecendented.
Where do you folks think intelligence, the ability to solve mental
problems, resides? If we talk about speeding up brains, well those
are just faster brains of the same intelligence. If we talk about
improving brains by giving them better software, i.e. better concepts
and insights, well that is just brains of the same intelligence who
know more. Somewhere you think that there is some special central
quality that remains fixed now, but will suddenly explode to higher and
higher levels once agents can hack their hardware.
Maybe, but maybe also what makes us smart is bascially what we know
and how fast we can think - maybe there isn't much more that matters.
If so, the ability to hack hardware will bring a big one-time jump in
raw speed, which will slow down into diminishing returns as most of
the easy speedups are found. And maybe if there are some qualitative
improvements possible in "basic intelligence", they won't be at all
easy to find.
Robin D. Hanson hanson@hss.caltech.edu http://hss.caltech.edu/~hanson/
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:35:46 MST