> Could you perhaps suggest some programming
> techniques that might in future produce the type of AI I have
> suggested, or is this not a realistic possibility?
If you want a detailed look at the programming issues, I'd suggest Eliezer
Yudkowsky's essay on the topic at:
http://www.pobox.com/~sentience/AI_design.temp.html His approach is a bit unconventional, but it looks promising.
Another approach would be to map out the general structure of the human brain, and duplicate the features of its components using evolved neural nets. The Robokenko project (see http://www.genobyte.com ) is doing some interesting work in this direction, although their goal is simply animal-level intelligence.
We don't really know enough yet to predict what approach will work best. However, I expect that we will see a lot of progress over the next few years. Computers are finally getting fast enough to support something more complex than an insect brain, and some of AIs long-promised goals are finally being reached. Hopefully, support for more advanced work will grow fast enough to let us take advantage of the new capabilities.
> I was interested to hear that some big project has been working on
> developing such an expert system for a decade. Do you have any more
> information on this project? Thanks for your input.
Sure. Its the Cyc project, and they've got a pretty good web site:
Their goal is to create a system that encodes all of the "common sense" knowledge required to understand your average encyclopedia article. They've produced some interesting results, but they still haven't really achieved their goal.
Billy Brown, MCSE+I