From: Ben Goertzel (email@example.com)
Date: Sat May 03 2003 - 21:23:30 MDT
> So to sum up: the idea of even excellent-level programmers being able to
> jump into this without deep understanding is not an idea we wish to
> contemplate or encourage at this point. Hence, need for book.
> Here's a page he wrote about this recently:
> Brian Atkins
I found that page very funny!
Especially the line
"Seed AI. It sounds like a cool idea, something that would look good on your
resume, so you fire off a letter to the Singularity Institute indicating
that you may be vaguely interested in coming on board. "
The idea that "Seed AI" would look good on anyone's resume, is pretty
hilarious!! My estimate of Eliezer's sense of humor has just been elevated
37.342 notches ;-)
I think that Eliezer's ideas about who should or should not become a seed AI
programmer, are tied in with his idea that it's necessary to FIRST create a
complete theory of AI friendliness, and only AFTERWARDS start making an
Artificial General Intelligence (AGI)
If you adopt the alternate point of view (shared by Peter voss and myself,
for example) that it's only through experimenting with real AGI systems that
we'll come to a real understanding of AI friendliness, then the story comes
out a little differently....
Nevertheless, I find something to agree with in Eliezer's comments. I have
found that most people who e-mail me and say they want to help out with the
Novamente project, aren't actually prepared to put in the amount of effort
that would be necessary to do so. This is partly because the Novamente
design is quite large and time-consuming to master, partly because the
Novamente project's organization is not all that newbie-friendly (our
tutorial documentation on the code is quite limited; there's no IT helpdesk,
etc.), and partly because building a thinking machine is just damn hard --
hard enough if you're giving it full-time effort, let alone if you're
squeezing it in after work and during lunch your. However, miraculously
enough, some folks have managed to learn enough about the design to
contribute meaningfully to the project even on a part-time basis...
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:42 MDT