From: Rolf Nelson (rolf.h.d.nelson@gmail.com)
Date: Sat Nov 10 2007 - 18:52:18 MST
On Nov 10, 2007 3:40 PM, Robin Hanson <rhanson@gmu.edu> wrote:
> Surely you can see these are very weak arguments in favor of any
> particular estimate.
>
Agreed, I was only addressing some common counterarguments than are weaker
than their speakers tend to think they are.
People like Kurzweil have the strongest arguments I've read for a Fast AI
timeframe (if you take out things like the infamous log-log chart), and
those arguments aren't very strong in an absolute sense. But they're still
stronger than the arguments I've read that we *won't* have Fast AI, so I
currently think Fast AI is more likely than not. If at some point I hear an
extremely persuasive argument for the Slow AI side, I'll change my mind,
probably from "Fast AI is likely and we should fund Friendly AI" to "Fast AI
is unlikely, but we should still fund Friendly AI just in case." A
persuasive argument for a specific "Slow AI" timeframe would have to address
progress in software, non-nanotech hardware, nanotech, and biotech
(including intelligence enhancement).
I'm not out to propagandize though, if you think Slow AI is correct than
obviously you should write unapologetically about that. Given the amount of
research you've already done, it's unlikely there are slam-dunk arguments
for Fast AI than you haven't heard.
-Rolf
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:00 MDT