From: Ben Goertzel (ben@goertzel.org)
Date: Fri Mar 01 2002 - 15:39:01 MST
> > Computers are too cheap and widespread, and so is programming
> > and cognitive science knowledge.
> >
> > AI can't be successfully regulated except via imposition of a
> > draconian, Luddite police state.
>
> Exactly, which is why I believe my proposal is germane. I doubt
> you could design a sufficiently comprehensive monitoring system
> to police AI development regardless of whether you wanted to be
> draconian or not. Accordingly, the proposal of identifying the
> best option and shifting vast resources towards that path is the
> most effective way to guide AI development.
The best option is to build a super-smart, super-Friendly AI fast,
before somebody else builds a nasty one.
I would be happy to see vast financial resources devoted toward
this project.
In fact, I would be happen to see even a modest amount of
financial resources devoted toward this project! I know from
experience however how difficult this is to achieve.
The fact is, the individuals and organizations
controlling the resources in our society
do not have a realistic view of future developments in technology,
and hence are very unlikely to fund such projects until success
(for good or for ill) has already been essentially achieved.
Hey, you work for Ernst & Young. When I was at Webmind Inc., I
tried to strike a deal with E&Y to supply AI components to their
knowledge management solutions. Guess what, it fell through due
to some corporate merger on their end (with Cap Gemini).
Had that gone through,
it could have generated profits to fund some of our real AI
development (which was "Friendly" in nature, although I don't
share all of Eli's views on goal system architecture). But, well...
just another example of how the individuals and organizations
controlling the resources do not value this sort of thing.
The only way the government funds AI significantly is by way
of military applications. Not the best route to Friendly AI
in my view.
-- Ben G
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT