Re: AI over the Internet (was Re: making microsingularities)

From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Sun May 27 2001 - 20:04:24 MDT


Spike Jones wrote:
>
> James Rogers wrote:
>
> > Spike is right; the existing computer infrastructure today is totally
> > inadequate for a "fast" takeoff via an Internet AI. Undefended computing
> > power is absolutely worthless if it is unusable as a distributed computing
> > resource for most intents and purposes....-James Rogers
>
> The person who pointed this out to me first over a year ago was...
> Eliezer. Perhaps I misunderstood what he said then, being as I
> am way outta my field of expertise. I dont even have a field of
> expertise. More like a lawn of pretty-good-ise. Or a weedpatch
> of mediocre-ise. {8^D
>
> I was asking back then if it would help in the progress of the
> Singularity to encourage massively parallel computing. That
> would cause jillions of desktop computers to be connected
> and be passing data on a regular basis. What I understood
> from his answer at the time is that it is practically irrelevant
> whether or not the personal computers are connected in
> that manner, because of the high inherent latency of those kinds
> of arrangements.
>
> Eliezer, did I misunderstand your comments from that thread? spike

Hm, I don't recall this thread well enough to remember. I would point
out, however, that this sounds like we were discussing the usefulness of
distributed Internet computing for *building* the AI, not the dynamic
*after* the seed AI gets started. Anyway, I really doubt I said that
distributed computing was irrelevant. What I may have said was that
Internet-distributed computing presents both obvious difficulties and
unobvious difficulties, which is why it wouldn't be such a great idea to
try to *develop* an AI distributed over the Internet, which was an idea
that I had briefly advocated before giving up on it.

-- -- -- -- --
Eliezer S. Yudkowsky http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence



This archive was generated by hypermail 2b30 : Mon May 28 2001 - 10:00:08 MDT