From: Stephen Reed (reed@cyc.com)
Date: Thu Jun 13 2002 - 16:07:21 MDT
On 13 Jun 2002, James Rogers wrote:
> > Cyc, which falls very short of AGI, is 1.4 million lines of translated C
> > code
> I don't think this is representative of the actual requirements though.
> I'd be willing to bet that the capabilities of the above system could be
> reproduced in an order of magnitude fewer lines of C code if
> re-architected from the ground up. That is a lot of lines of code for
> something that doesn't do that much in an algorithmic sense.
Yes, if a deductive inference engine and symbolic knowledge representation
repository were essential for Seed AI, then I agree that less than 20K
lines of java code would be required, especially if a relational database
were the object store. It would compare very poorly to Cyc but might be
enough for Seed AI, given the assumption that shallow deductive inference
would be the requirement. A substantial portion of the Cyc code base is
motivated by computational linguistics, and I am leaving that
behavior aside.
> > So I believe that the creation of Seed AI is far beyond the capability of
> > a single individual, and that belief is my motivation for working at
> > Cycorp.
> This is a flawed belief in that it appears to be based solely on the
> premise of your experience with Cyc. If you don't think the Cyc
> architecture is directly pertinent to AGI (and many don't), then it
> becomes a null data point.
Well yes, I agree that my Cyc experience can be discarded for those who
doubt our approach. Yet add my experience to that reported by other AI
projects.
> My personal meaningless guess with respect to the ceiling for the number
> of lines of compiler code required to generate an AGI is about 50,000.
My question to you then would be: Given a budget of 50K LOC, just what
behaviors would you require the system to have? Or if your base system
merely accepts knowledge for the next layer up, how many person-years of
effort at that next layer would be required to achieve Seed AI, and what
behaviors would be taught to the system with this budget.
All the implementers on this list gain from a discussion of the
essential Seed AI behaviors, regardless of how we achieve those behaviors
in a software system.
Is reading important? What kinds of machine learning are required? Does
automated reasoning require justifications/explanations? What are the key
attributes of a goal? ...
-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