Re: META: Killthread; (Re: Edge.org: Jaron Lanier)

From: Perry E.Metzger (perry@piermont.com)
Date: Sat Nov 29 2003 - 09:31:17 MST


Tommy McCabe <rocketjet314@yahoo.com> writes:
> Technology, at least under my definition, covers a
> very wide range of topics, and simply lumping
> something in with "technology" doesn't make it
> relevant to SL4. IBM, Intel, AMD, etc., are doing a
> very nice job of making chips, and unless AI somehow
> requires either specialized types of hardware, or very
> thorough knowledge of the hardware it is being
> programmed on, there is no need to waste time
> discussing it.

I couldn't disagree more strongly. It is hardly clear that
conventional hardware is conducive to the sorts of things that are
needed here -- or conventional programming paradigms.

The position that we can ignore the hardware is one that betrays a
lack of long term perspective, IMHO. (The position that software can
be ignored among hardware designers creates similar tunnel vision. Oh,
for a dollar for every processor designer who thought virtually
indexed caches are a good idea, or for every designer that assumed he
knew what the driver writer would want without asking one...)

BTW, it is also hardly clear that conventional hardware will "last"
long enough. The end of Moore's Law comes apace, and we'll have to
start dealing with parallelism whether we like it or not.

> If it isn't broken, don't fix it, and
> don't spend valuable time discussing it. Quote from
> Staring into the Singularity- "Ever since the late
> 90's, the Singularity has been only a problem of
> software."

Also, largely, something I can't agree with. I'm a strong AI
proponent, but merely because all hardware is Turing Equivalent
doesn't make it equally convenient from an engineering perspective --
and it is also hardly clear that our solution will look like software
in a conventional sense, either!

(In theory, of course, Cobol and C are of equal power -- but I
challenge you to write an operating system kernel in Cobol.)

> The hardware companies can handle the
> problem of making fast chips- but we need the code to
> make the chips become a Friendly Seed AI. And that's
> where SIAI comes in.

And why would one expect SIAI would come up with a solution, or that
it could be done purely with conventional hardware? (Repeating, just
because a variety of systems are Turing Equivalent doesn't make them
all equally convenient for one's work.)

Perry



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:43 MDT