From: Stephen Reed (reed@cyc.com)
Date: Mon May 20 2002 - 12:51:51 MDT
On Mon, 20 May 2002, ben goertzel wrote:
>> Our system is so large, and our projects so pressing, that our day
>> to day focus is almost entirely incremental.
> This approach worries me.
>
> I am not sure that this kind of "hill climbing" approach to AGI can work
> (incremental improvement of Cyc being essentially the hill-climbing
> algorithm applied in "AI design" space ;)
>
> You may well incrementally improve yourselves into a local optimum in AI
> design space, I think...
I like the metaphor and appreciate the comment.
Cycorp does however have the opportunity to evaluate the situation and add
entirely new features, for example: the planner and blackboard which did
not arise incrementally.
A solution, in my opinion, to the hill climbing local optimization of the
design/effort space, is a roadmap that covers enough behaviors so that
local maximums are surpassed by global maximums. For example we could
improve deductive reasoning for a very long time, but if Induction,
Abduction, Estimation, Prediction, Modeling, and Problem Solving are
desired Cyc activities, then we will allocate resources to these other
behaviors, even to the detriment of progress in deduction. At the moment
we have no such roadmap, but I am influencing our management in that
direction, and helping to write government proposals that mention them.
-Steve
--
===========================================================
Stephen L. Reed phone: 512.342.4036
Cycorp, Suite 100 fax: 512.342.4040
3721 Executive Center Drive email: reed@cyc.com
Austin, TX 78731 web: http://www.cyc.com
download OpenCyc at http://www.opencyc.org
===========================================================
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT