RE: AI hardware was 'Singularity Realism'

From: Keith Henson (hkhenson@rogers.com)
Date: Sat Mar 06 2004 - 15:17:46 MST


At 12:22 PM 06/03/04 -0500, you wrote:

> > >Of course, if I speak from my gut, I must admit don't even think it
> > >would take a Manhattan Project style effort. I think I could get
>there
> > >in 2-3 years with an investment of $20 million for staff and
>hardware.
> >
> > If you can make a strong case I think it could be funded without much
>of a
> > problem.
>
>Keith, I am curious what funding sources you're thinking of.

I know at least one computer savvy person who could write a check that large.

>The major science and technology funding bodies in the US are very
>tightly tied into the "narrow AI" research programme and very skeptical
>of radical approaches to AGI. This is in spite of the fact that they've
>spent hundreds of millions of dollars on narrow AI programs (Cyc being
>the flagship example ;-) without obtaining dramatic returns either
>scientifically or economically.

That's not surprising considering how much computational power biology
lavishes on the problem. Have you ever looked up the MIPS rating of a retina?

>So, if by a "strong case" you mean a case that will convince AGI
>skeptics such as the folks at the National Science Foundation -- I guess
>this essentially means "something like young-child-level human
>intelligence has been achieved and what remains is teaching of the
>system and refinement of the algorithms."

I would not need that level, but you need to sell me that you have a viable
approach before I stick my neck out and try to convince people with money.

>I have no doubt that once we
>get to this point, ample funding will be available. Also, by this
>point, the political issues associated with AGI will become important --
>the Luddites will start paying attention....

Possibly.

>The problem is doing the work to get our AGI design (Novamente)
>implemented and tested and tuned to the point of skeptic-convincing,
>young-child-level human intelligence. Getting the funding for this work
>is the problem.
>
>Fortunately, the Novamente project *is* funded; however, the funding is
>for commercial applications of the partially-complete codebase, which
>allows us to do some fascinating and important application work, and to
>push the AGI project along slowly, but not to proceed rapidly toward AGI
>in the way that we would attempt if we had pure AGI research funding.
>
> > But I don't think anyone right now has a clear idea of what it would
>take
> > to build an AI by other means than a simulation of a human brain.
> > (Correct
> > me and send pointers if someone *does* know.)
>
>Obviously, I think I do have a clear idea in this regard. However, our
>AI design has not been published or presented on the Internet. We plan
>to publish a lot of material on Novamente sometime in the not too
>distant future, probably early 2005 it looks like. Writing up stuff
>like this in a quality way is time-consuming, and time spent writing
>needs to be balanced against time spent doing.

We might consider spending an hour or so in chat. If nothing else, you
could salvage the chat log for a written presentation material.

I happen to be a bit skeptical that the hardware is up to the task based on
arguments by Hans Moravec, Ray Kurzweil and others. In the long run this
is not a problem since hardware equal to the task is less than a human
generation away. If you have a radical approach that would allow cockroach
level hardware to generate superhuman AI level performance, I would sure
like to know what it is.

Keith Henson



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:46 MDT