Fw: Failure of AI so far

From: Mark Waser (mwaser@cox.net)
Date: Fri May 23 2003 - 20:01:20 MDT


Dani Eder said:
> I would expect as computing power grows, the
> programming task gets easier, since you can implement
> simpler but more compute-intensive methods. In the
> limit of sufficient computer power, you can model at
> the neuron level, then repeat 10^11 times.

You could, but then you would have absolutely no idea of what your AI was
thinking and no way to ensure it's friendliness. This is the ultimate
nightmare case for most of the people on this list - - the hardware improves
fast enough that someone does something like that before someone succeeds in
developing a Friendly AI.



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:42 MDT