From: James Higgins (jameshiggins@earthlink.net)
Date: Thu Jun 27 2002 - 17:22:16 MDT
At 07:01 PM 6/27/2002 -0400, Smigrodzki, Rafal wrote:
>Ben Goertzel [<mailto:ben@goertzel.org>mailto:ben@goertzel.org]
>
>
>History shows that throwing money at a problem is not always the best way
>to solve it!
>
>This is why these days, major software firms tend to effectively outsource
>most of their software R&D to startups -- acquiring those startups that
>have produced something clearly valuable.
>
>### Yes, the basic research takes time, genius, and luck to succeed. But
>once a good proof of principle experiment is successfull (fission in a
>small chunk of uranium), the task of transforming it into a useful product
>(the nuclear weapon) might require only a massive engineering and
>sclaing-up effort. I do not know enough about AI (it's fair to say I know
>almost nothing) - so instead let me ask you: Do you think that there will
>be a phase in the AI project (perhaps at the high-infrahuman level) where
>further improvements will become a matter of computational and
>organizational muscle applied to the principles you develop?
Of course there is. Once the infra-human AI is up an running. At this
point most of the effort shifts to teaching and growth and intelligence
enhancement is dependant on computing resources.
The DOD could provide many times the computational resources virtually
anyone else could afford to provide. And in a massively parallel effort
they could have literally hundreds of butts in seats teaching the AI around
the clock.
James Higgins
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT