Re: State of the SI's AI and FAI research

From: Randall Randall (
Date: Wed Feb 16 2005 - 15:35:39 MST

On Feb 16, 2005, at 4:04 PM, Matt Arnold wrote:

> Eliezer,
> Your comments remind me very much of me. I find saying things like
> "proportion the strength of convictions to the strength of the
> evidence" is easy for me to say, but not so easy to put into practice
> consistently. Seeking out the strongest contradictory hypotheses and
> alternative explanations, and attempting to understand their
> perspectives, is a daunting effort. I hate to admit it, but it's been
> too much for me to keep up day in and day out.

Indeed, the only way to get through each day is to
memoize the old decisions for any input which is
similar, except when there's extra time and energy
to explore the solution again. However, this only
works for decisions that are "really" the same
decision, and not those that are very different but
seem similar on the surface.

For the kinds of problems that Eliezer is tackling,
however, there appear (from this outsider's view)
to be many decisions which look similar to those not
directly in the field, but which are not similar
enough to have old results apply usefully.

Such a strategy that works for most applications,
then, isn't so useful when doing something that is
probably truly new.

Randall Randall <>
Property law should use #'EQ , not #'EQUAL .

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:50 MDT