Re: [sl4] Re: More silly but friendly ideas

From: John K Clark (johnkclark@fastmail.fm)
Date: Mon Jun 30 2008 - 11:26:52 MDT


On Mon, 30 Jun 2008 "Stuart Armstrong"
<dragondreaming@googlemail.com> said:

> even the crudest of cut-offs would solve
> an infinite loop problem

But when do you cut it off? Perhaps in one more second you will get your
answer, or perhaps in 2 seconds, or perhaps in 10 billion years, or
perhaps never. There is no way to tell.

> Please give a specific example of something
> a goal orientated AI would fail at, because
> of Godel-type considerations.

There are an INFINITE number of such problems, but Turing proved there
is no way in general to determine if any specific problem is of that
nature or not, and that can cause trouble. One possible example of an
imposable goal is “Prove the Goldbach conjecture to be true or find a
counter example to show it is false”. And if Goldbach is not one of
those true but un-provable statements you can be certain there are an
infinite number of similar statements that are.

  John K Clark

-- 
  John K Clark
  johnkclark@fastmail.fm
-- 
http://www.fastmail.fm - Faster than the air-speed velocity of an
                          unladen european swallow


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT