Re: Superintelligence Software

Brian Atkins (
Sat, 08 Aug 1998 18:23:03 -0400

Technotranscendence wrote:
> At 11:07 AM 8/8/98 EDT, Jay <> wrote:
> >Yes, this to has bothered me. The advances in hardware seem predictable and
> >on track, but given the bug ridden examples of SW we now see, it looks like
> >the SW problems might be harder.
> >Do any SW gurus on the list have any Morvec like graphs indicating the
> >progress in SW algorithms? Anyone that uses windoze should have their doubts.
> >I think that some sort of self replicating/self generating SW is needed with
> >an evolutionary correction/redundancy built in will be necessary. Any
> >progress in this area?
> I think genetic programming will lend itself to this treatment.
> Daniel Ust

Agreed, it just seems beyond human ability to create software from scratch to be able to do everything we've evolved to be able to do. Some AI teams (such as in Japan) are now experimenting with evolving subhuman AI creations...

I think the only path to a human-level AI will be to evolve it inside of a totally realistic VR simulation, which mimics how we came to be, just a whole lot faster (hopefully). This was demoed in Egan's Permutation City.

The future has arrived; it's just not evenly distributed.
                                                       -William Gibson
Visit Hypermart at for free business hosting!