From: Eliezer S. Yudkowsky (firstname.lastname@example.org)
Date: Sat Jan 29 2005 - 08:34:39 MST
Eliezer S. Yudkowsky wrote:
> ... would suggest that I
> try writing code that executed with a bound of ten thousand sequential
> serial operations, to get a feel for how restrictive that limit was.
Come to think, I've been in the programming game long enough to have
actually done this. Back when computers were slower, I ran up against the
30,000 tick bound in the MOO programming system. I was a timing freak in
my early days. I would take other folks' MOOcode that had formerly taken
ten thousand ticks on average and would occasionally cause runtime
exceptions when it ran out of time, and rewrite the code with an improved
architecture so that it would execute in less than a thousand ticks. Yes,
I know today that this is over-optimization and that I was teaching myself
poor programming habits. I was fourteen at the time.
So a thousand-tick restriction is not as restrictive as all that in
ordinary programming - though a tick of MOOcode translates to considerably
more than one assembly-language instruction. Still, AI coding is not
ordinary programing, and I don't think anyone would dare to hypothesize -
did they not have the example of humans before them - that the Sacred Grail
of Intelligence might be doable at 200Hz.
People confuse ignorance with difficulty. Intelligence feels extremely
hard because they don't know how to do it. Obviously, a problem so
difficult as intelligence cannot be solved with 200Hz processors.
-- Eliezer S. Yudkowsky http://singinst.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Sun May 19 2013 - 04:01:09 MDT