RE: AI timeframes

From: Ben Goertzel (ben@goertzel.org)
Date: Fri Apr 09 2004 - 05:31:27 MDT


> Tomaz Kristan wrote:
> > I have seen enough history to conclude, that the US
> government has no
> > other choice, than to start a Manhattan size SAI project
> [...] And I
> > have no problems with this scenario.

IMO, the AGI problem does not NEED a unified project of the size of the
Manhattan project. Research software is not a large-team enterprise. I
learned that the hard way!

If a lot of money were going to be thrown at the problem, the most
intelligent approach would be to

1) buy a shitload of hardware and maintain and upgrade it well
2) fund a number of independent, autonomous small teams working on their
own AGI approaches, using the hardware farm

Also you have to consider the fact that the government R&D establishment
(particularly DARPA) has heavily bought into a non-workable approach to
AI (expert systems, formal logic as the central principle, Cyc, etc.).
This means that, unfortunately, it's quite possible that a
government-funded Manhattan Project for AI would wind up being
Cyc-centered! The affinity of the government for this particular
species of non-AGI-friendly narrow-AI technology means that a workable
"Manhattan Project for AI" is not as imminent as you might think. To
create such a thing would require someone to do some very clever
politicking within the government research establishment. Which of
course is not impossible...

- Ben Goertzel



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