Re: Seed AI (was: How hard a Singularity?)

From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Mon Jun 24 2002 - 17:02:01 MDT


James Rogers wrote:
> On Sun, 2002-06-23 at 11:47, Eliezer S. Yudkowsky wrote:
>
>>In my experience, your instincts are correct; this doesn't apply to AI, but
>>applies to everything else. Excessive complexity usually arises from a lack
>>of true understanding of the problem domain, and simple systems with good
>>architectures are usually far better than complex ones. Except in AI, where
>>the problem is far more complex than your brain "wants" to think - in part
>>because of all the complexity of thinking is invisible to you, and in part
>>because usually humans just don't deal with things that complex.
>
> This is a bizarre comment, and I am having difficulty interpreting it in
> a way that makes sense. Complexity does not have a subjective
> interpretation in proper engineering, and it doesn't follow that AI is
> in some special category of problem. Bad engineering practice reflects
> on the engineer, not the problem. Or to put it another way, regardless
> of how complex the problem actually is, how do you know your perception
> of the complexity is even remotely correct without going through the
> rigorous theoretical normalization required to safely make that
> assertion? This is a particularly relevant point if we are premising a
> problem space that we are hypothesizing to be complex as such things go.

An AI can have a very elegant design and still be far more complex than what
human programmers are used to. Overcomplexity and oversimplification are
relative to the problem, not relative to what strikes humans as being a lot
or a little complexity. The natural complexity of AI is a lot higher than
what programmers are used to, so in an absolute sense a "just-right AI" is
probably more complex than most "overcomplicated" programs. Yet an AI built
"just right" might still have in all other ways the signature of good
design, rather than bad design.

"When they discovered that reality was more complicated than they thought,
they just swept the complexity under a carpet of epicycles. That is, they
created unnecessary complexity. This is an important point. The universe is
complex, but it's usefully complex."
        -- Larry Wall
        http://www.oreilly.com/catalog/opensources/book/larry.html

-- 
Eliezer S. Yudkowsky                          http://intelligence.org/
Research Fellow, Singularity Institute for Artificial Intelligence


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT