From: Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Date: Wed Dec 08 1999 - 12:56:01 MST
Robert J. Bradbury writes:
> Gentlemen, as Greg pointed out in August, I've demonstrated the
> Alpha Male Syndrome (AMS) adequately (in my discussions with John)
> for the 1999 archives. Its probably advisable for further exhibitions
> to await either the coming millennia or some hard data on which
> the beating on one's chest is justified. (If you think about the
I think reasoning clauses and hand-code a true AI is now fairly firmly
disproved. I like LISP as much as the next person, and personally
think some of the biggest and most robust systems even have been
written in it. Heck, I use XEmacs, and I love it despite all its
failings. God knows, I was even a firm believer in AI in mid-80s
myself. I simply didn't knew better back then.
However, I now do not see any alternative in stealing from the design
of life, either by blind copying the blueprints, or (better)
harnessing the metamethod (evolution) itself.
Nonbrittleness, massive finest-grain parallelism and an astonishing
economy of resource utilization are not possible to achieve
otherwise. This comes at a price, of course: the evolved designs are
not understandable by a human. A nightmare of coupled autofeedback
loops do not appeal to the orderly engineer's mind. However, as long
as it outperforms anything the engineer can produce, I do not really
care.
If it thinks, do we really have to know in detail how and why it
thinks? And if we _really_ have to, we can really find out later.
Weird? Sounds like a badge of honour to me.
> list en-toto, you can imagine the number of --
> "See, I told you so!"s
> that people are saving up...)
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:06:01 MST