RE: Prediction.

From: Ben Goertzel (
Date: Tue May 14 2002 - 22:41:58 MDT

> More basic
> science in the fields of computer science, nanotechnology, and AI has come
> out of IBM than any other single place in the world. IBM has the
> resources
> to have 50 Novamente projects going on in secret.

I happen to be AT IBM right now, in Yorktown Heights. I have been meeting
with a number of their computer scientists, in the context of a conference
on "autonomic computing".

The fact is, they laid off most of their AI staff in the mid-90's, keeping
only those who were willing to work on very narrow-focused AI projects.
They have a handful of staff working on AGI-related stuff, individually and
without an overall plan, with almost no funding ... most of them spending
half or more of their time on non-AGI stuff.

The few staff they have working on AGI stuff, individually and in their
spare time, are not too happy with the situation.

Sure, all these staff scientists could be lying to me, or could be unaware
of the top-secret AGI projects going on, but it's amazingly unlikely. The
public record of their hirings and layoffs and quittings supports the story
told by their staff.

>You think the Deep Blue
> II beating Garry was a parlor trick? I happen to be a chess master and I
> can tell you that what happened there was amazing. It is often sited the
> billions of moves per second that DB was checking, as if it beat Garry by
> pure brute force. Garry went into that match knowing that you can't beat
> Garry that way. Garry threw a fit and accused the IBM team of cheating.
> Why? Because DB thought beyond the board, simulating Garry himself, and
> used a psychological ploy to trick Garry into making a bad move and losing
> the game. There was much more going on in that program than chess.

I know some of the people who worked on Deep Blue. What you say is not at
all accurate. Deep Blue used min-max search with alpha-beta pruning, and a
clever heuristic evaluation function created and tuned by some chess
experts, all implemented on specialized hardware.

> And I
> wouldn't count Bill Gates out either. M$ has it's own AI research
> department which they claim is doing GOFAI. It probably is doing some of
> that along with the real work. Nor would I discount Japan, Australia, and
> fifty universities around the world. Underground AI is BIG BIG
> BIG!

You are simply dead WRONG, WRONG, WRONG! M$ does not have a significant AGI
research program, and neither do any major US universities. You can see
this, for example, by studying M$'s hiring patterns. they have hired
extensively in areas like graphics and bayes nets but not AGI. As for
universities, all uni research is funded by gov't grants, and the NSF and
other gov't programs have not funded AGI research significantly for decades.

Australia has no significant AGI research, at any of their major
universities,not even the ANU.

Japan's biggest AI effort is at ATR, but they reorganized a couple years
back. Hugo de Garis moved to Belgium and Katsunori Shimohara's Human
Information Processing Lab was renamed and reorganized; advanced AI is on
the outs there, they're mostly into moderately-intelligent robotics...

> I like
> Kurzweil's charts because they are data points and not as subject to
> conservativization. But his formula is purposely biased to the future. I
> think he knows it's coming in 2005 but doesn't want anyone else to know.
> His prediction of 2005 - 2030 is to make people think it is longer while
> preserving his ability to not be proven wrong. He wants to a let people
> that would be in favor of the Singularity know that it is coming but he
> doesn't want to alert the enemy how soon. What would happen in January of
> 2005 if CNN, started broadcasting that experts predicted the
> Singularity in
> four months? Mass panic.
> Prediction: April 28, 2005.

Not to be too blunt about it, but I think either you're half-nuts or you're
just playing around ;>

It's not your prediction in itself that makes me say this, it's the
dead-wrong reasoning underlying it.

The pattern of top-secret AGI research that you cite simply does not exist.
if there are such top-secret programs they are not at M$, IBM or
universities, they would have to be deep within the NSA or CIA or
something... and there is no evidence to suggest this at all (e.g. no hiring
of AGI PhD's by such agencies).

I think a Singularity by 2010 is *possible*. If Novamente or A2I2 or
Mentifex or NARS or DGI some other AI project goes *really well*, this could
happen. Not impossible, and I hope it happens.

But given the current state and pace of development of the various known AGI
projects, and the incredible unlikelihood of advanced unknown projects of
this nature, 2005 is incredibly unlikely.

-- Ben G

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:38 MDT