From: Marc Geddes (marc_geddes@yahoo.co.nz)
Date: Wed Nov 03 2004 - 21:13:32 MST
--- Eliezer Yudkowsky <sentience@pobox.com> wrote:
>
> 56 people donated a total of $8,071 during the Today
> and Tomorrow campaign.
> That's more people than donated during all of
> 2003.
>
> The country of Denmark is now represented on the
> monument. A toast to Denmark!
>
> Finland continued its steady march toward galactic
> domination.
>
Sitting around fantasizing about building monuments
post-Singularity won't get us anywhere. We're
currently stuck on the other side of the Singularity -
and what a banal, brutish existence it is!
I estimate that any serious AGI project (including
Sing Inst) needs an income level of at least $US 300
000 per year to 'ramp up' to full-scale operations.
6-8 full-time team members, who would need to be paid
at least the 'entry level' salary of $US 25 000 per
year. That's $US 150 000 per year for a team of 6.
Then about the same amount for office space, hardware
etc. So at least $US 300 000 per year. I’m glad
progress is being made, but I still see no sign that
Sing Inst is yet near that sort of cash flow.
Worse, AGI theory seems to be getting more
complicated. There is more to CV than to earlier
versions of FAI theory, and it would appear to my eyes
that for every AI problem Eliezer is solving, two more
are popping up. My suspicion is now overwhelmingly
strong that (a) CV cannot be calculated by a
Singleton, (b) that completely selfless AGI is
impossible, and (c) that real-time general
intelligence without sentience is impossible. If I'm
right, even CV is not the last word, and yet another
major theoretical change is yet to be made by Eliezer.
Incidentally, assuming a successful Singularity,
post-humans are right now poring over the SL4
archives. They are no doubt of great interest to
post-human historians. Interesting really. Either
I'll go down in the history of the galaxy as a total
clown, or I'll be remembered forever as the guy that
actually challenged the great Yudkowsky and made
several key conjectures that were later proven
correct. Moral of the story: post to SL4 and the
post-humans are watching. The whole galaxy is reading
these archives. (Assuming a successful Singularity of
course).
Back on topic: The increasing difficulty, the lack of
funds and so on. Let's give Eliezer the benefit of
the doubt and assume that he now understands all the
necessary conceptual fundamentals of FAI theory (which
I'm pretty sure he doesn't). Even so, formalizing all
that should take at least 15-20 years. I base this,
as I said in an earlier post, on the length of time it
took other geniuses to move from principles to maths.
And the level of difficulty is likely to be 'gut
busting' - probably the sort of math’s required is at
the 'Riemann Hypothesis' level of difficulty. You
shouldn't kid yourself - even if everything goes
right, I don't see anyone getting to FAI in less than
15 years. So that's 2019 at the earliest.
It gets worse. The existential threats may start
popping up earlier than I thought. I was reading some
stuff from the Center For Responsible Nano (Mike
Treder's think tank). I originally thought nano
wouldn't be threat until 2030 and beyond. But now
Treder and others are pretty sure nano's gonna be
hitting in force as early as 2020. So the risks are
going to start going up after 2020. This date is just
after the shortest time I think it's possible to get
to FAI ;)
Eliezer's best years might soon be behind him, as he
said. The AI problems are popping up faster than he
can solve them, and he could soon start slowing down
with age. There's no funding. FAI theory keeps
getting more complicated. Formalizing conceptual
principles takes 15-20 years. And existential risks
from nano could be as little as 15 years away. It's
not looking good at all.
So: Only a small band of dedicated Singulatarians
know what's going on, long odds, certain death if we
fail fate of the universe at stake. I love it :D
=====
"Live Free or Die, Death is not the Worst of Evils."
- Gen. John Stark
"The Universe...or nothing!"
-H.G.Wells
Please visit my web-sites.
Sci-Fi and Fantasy : http://www.prometheuscrack.com
Mathematics, Mind and Matter : http://www.riemannai.org
Find local movie times and trailers on Yahoo! Movies.
http://au.movies.yahoo.com
This archive was generated by hypermail 2.1.5 : Tue Feb 21 2006 - 04:22:48 MST