From: Charles D Hixson (charleshixsn@earthlink.net)
Date: Fri Aug 25 2006 - 13:32:30 MDT
Eliezer S. Yudkowsky wrote:
> J. Andrew Rogers wrote:
>>
>> Community-based AI research is the idea that having a clue is an
>> emergent property of having enough people without a clue in one
>> room. And I use "emerge" in the "magic happens" sense here. I would
>> make the observation that most productive AI research seems to be
>> done by people who have shown little interest in community-based AI
>> research, which would suggest that the "community" aspect is quite
>> irrelevant to their progress. It is not as though there are not
>> plenty of community-based AI research projects in existence.
>
> The Japanese Fifth Generation project was exactly this sort of
> project, only better funded. I wonder what went wrong. ...
The Japanese PR claim was that it was that sort of project. Actually it
was an attempt to bootstrap the Japanese computer industry and support
their mainframe builders. Nothing went wrong. You can argue that they
didn't get their money's worth, but it was a pork-barrel project, so
that's not an arguement.
Actually, they were starting too soon to achieve the kind of
breakthrough they were announcing they were trying for...and it would
still be too soon. At this stage what should be funded are lots of
relatively small projects...say 8 college kids & their parent's garage
size. Aim for "just graduated" and "dissertation work" to stretch your
money. DON'T pay top dollar, aim to build skill sets. Require that the
government (or other investors) get 50% of all profits & of all IP (all
three kinds). Provide a "business services" unit to oversee compliance
with relevant laws. (Presume that EVERYONE hired under this program is
a developer...so all overhead is born by the investors, and that means
business licenses and tax law compliance as well as everything else.)
After a year, demand a proposal for a "demonstration" that they think
that they'll be able to do in a year, then allow them two years to build
it. If they fail, all of their work becomes Open Source (BSD or GPL?)
and funding is cut in half. And they get a year extension.
That kind of thing. This shouldn't be expected to yield any
breakthroughs, but LOTS of skill development and SOME technical
progress. And a few modest successes.
I can see also funding around 1/3 as many projects at 3 times the
funding, but these would need to be more carefully selected, and would
need to show something before starting. And possibly 1/9th as many
projects at 9 times the funding. etc. At each step the criteria for
selection would need to be more than proportionally stronger. The idea
here is not to create a monopoly, but a strong base. Monopolies are the
WRONG direction.
OTOH, would anyone here really want to merge their project into one that
was run by the government? Defense contractors end up spending a lot of
time lobbying Congress for support, and the government places insane
restrictions on their businesses. And they tend to favor a very few
large companies that they can, essentially, totally control. (I'll
grant that the original transformation of the defense contractors was
slow, but new ones are expected to fit into the existing framework
immediately.)
FWIW, and frustrating as it is, I suspect that underfunded and free is a
better state for technical and philosophically correct development. I
also expect that the government is already funding lots of AI
projects...and that it's those that need to be beaten if one desires FAI
as opposed to UFAI.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT