On 10 Nov 2006 at 13:48, Eric Baum wrote:
> I am arguing that if you find a code that solves a range of naturally
> presented problems, that is so constrained (for example by being
> extremely concise) that the only way it could have existed
> is if it is based on an ingenious abstraction hierarchy with modules
> exploiting real structure in the world and reused in different
> contexts to solve different problems, then it will continue to solve
> most new problems,

Ok, that seems pretty obvious, at least for anyone who accepts
the basic premise that the real world is highly ordered. I don't think
you're saying anything controversial there, though people who don't
like the idea of AIs being based on serial-looking code will probably
nitpick 'range of naturally presented problems' in the hope of
demonstrating that only fuzzy/connectionist/massively parallel
code can solve nontrivial challenges (I'm not one of those people).

Generalising to this level isn't really saying anything useful though,
as you've pushed all of the challenge into finding that code/module
set/abstraction hierarchy in the first place. Clearly you have done a
lot of work on that challenge, but that may not be apparent to people
reading these sweeping generalisations. Solid statements about how
to handle uncertainty and lossy compression, however concise, would
help make this descriptive (in the sense of definitely selecting a narrow
subset of possible designs; overly general bits of AI wisdom suffer from
the problem that virtually any researcher can claim compliance with
them with a bit of judicious rephrasing and repackaging of whatever their
current ideas are). Though I'll understand if your response is another
'just read the book' - reiterating one's ideas can get tedious. :)

Michael Wilson,
Director of Research and Development, Bitphase AI Ltd
Web demos page: http://www.bitphase.com/apex


-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303

Reply via email to