Re: Investing in FAI research: now vs. later.

From: Rolf Nelson (rolf.h.d.nelson@gmail.com)
Date: Mon Feb 11 2008 - 19:05:24 MST


> They will at least try to make the AI safe, just as weapons
> researchers try not to blow themselves up. Whether they succeed or not
> is another matter.

Eurisko didn't, nor (to my knowledge) did they have any safeguards that, had
the project succeeded in creating an AGI, would have prevented one reseacher
from typing "give me rulership of the world, moo ha ha ha" on the keyboard
while the other researcher was taking a bathroom break. It really wasn't on
their radar screen.

On the plus side, the Manhattan project succeeded with researchers who did
take basic safety precautions, rather than just place enriched uranium in a
pile until it blew up.

It's possible (but not, unfortunately, anywhere near guaranteed) that
Competence in Taking Safety Precautions will correlate strongly enough with
Competence in Building an AGI that the first team to build an AGI will be
competent in both areas.



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:02 MDT