From: Gordon Worley (redbird@rbisland.cx)
Date: Tue Jul 31 2001 - 23:44:25 MDT
At 2:06 AM +0200 8/1/01, Joaquim Almgren Gāndara wrote:
> > Very true. But if we can get an AI to at least a 1.0 level, then
>give it
>> sufficient processing power so that it is much, much faster than a
>human,
>> it will progress on its own. Because it will make tiny advances
>over time
>> (on it's time scale) which would lead to 1.1, 1.2, 1.5, 2.0, etc.
>If 1.5
>> takes more processing power, it could slow itself down some but
>improve the
>> quality of its thought.
>
>I don't quite understand this. Why place the threshold at 1.0? What
>evidence is there indicating that the average human (plus AI
>advantages such as codic cortex) is smart enough to progress to higher
>levels of intelligence? Why not someone half as smart (0.5) or 1.5 or
>7.0 or any other arbitrary number? I think you're being
>antropocentric.
The reason for this is simple. First off, we have Ben, so we know
that human level intelligence is enough to work on AI algorithms and
make meaningful improvements. I'm willing to bet that, given enough
time, Ben could keep making improvements, though as time goes on it
will be harder. Of course, for an AI this isn't a problem since it's
dynamic, but I digress. Secondly, how many dumbasses (yes, I think
this is now a technical term ;-)) do you know that can improve an
algorithm by any method other than getting lucky? Personally, I know
of none who can do that, but then I don't know that many. But, I'm
willing to make another bet that infra human intelligence is not
enough to make useful improvements to algorithms within a reasonable
amount of time (if ever).
But, because I think we are getting way to anthropomorphic, I'll
bring up that measuring AI in terms of human brains probably won't
work. If I were building an AI, first I'd want it to learn specific
knowledge about computer science and then, as it improves itself,
teach ver about more.
-- Gordon Worley `When I use a word,' Humpty Dumpty http://www.rbisland.cx/ said, `it means just what I choose redbird@rbisland.cx it to mean--neither more nor less.' PGP: 0xBBD3B003 --Lewis Carroll
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT