From: Eugene Leitl (eugene@liposome.genebee.msu.su)
Date: Wed Jul 15 1998 - 08:17:47 MDT
Hara Ra writes:
> The bigger, faster, louder is better meme has been around for a long long
> time.... If you take a dog, upload it and run it at 10^6 that of realtime,
> you have a very fast dog, but it still is a dog. I think we have a lot to
> learn before a SI becomes feasible.
Agreed. A dog is a dog is a dog, but a god is just a few exponential
steps off. If according evolutionary trajectory is kinetically
accessible: instant SI, just add ZOPS.
> I note these days with dour pleasure that the hardware has gotten so much
> ahead of the software that most folks like 266 Mhz $1000 Wintel boxes so the
> PC more (everything) to run (huger and huger) bloatware trend is less now.
In terms of statespace velocity, the hardware has not progressed all that dramatically.
The reason we _think_ the hardware has progressed so much beyond the
software is one of the current tradegies we have to live with. Why are
people so intent on censoring their own worldview? Everything which
even distinctly smacks of von Neumann is wrong, Wrong, WRONG.
> Someone with better data that I have (John Clark?) might think about how many
> machine cycles are equivalent to evolution of human beings from
> archaeobacteria,
> as this might provide a clue as to just how fast the hardware must be to
> create
> an SI from evolutionary principles. I sort of doubt it is less than say 10^50
> operations....??
1) We're not archae, and we understand our hardware more or less
thoroughly (or at least are just a few years off that stage). We
don't have to retraverse all these GYrs.
2) we can write our fitness function. Nothing prevents us from
starting off with realtime robotics and Turing test performance,
while optimizing for SPEED.
3) evolution speed in EHW fitness space is nothing like precambrian
ursoup simmering forever. Electronically excited Genotype2phenotype
generation iteration takes microseconds, not decades.
To sum it up, if we do it right, I think we have excellent chances to
leap beyond our current status quo in a few bounds. Let's.
'gene
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:49:21 MST