Some Challenges And Grand Challenges For Computational Intelligence

Printable Version

by	Edward Feigenbaum

The Turing Test is a very ambitious Grand Challenge. The "Feigenbaum Test" is more manageable: focus on natural science, engineering, or medicine with conversation in the jargonized and stylized language of these disciplines. There are two other grand challenges in achieving Computational Intelligence: Build a large knowledge base by reading text, reducing knowledge engineering effort by one order of magnitude; and the "Grand Vision": distill from the WWW a huge knowledge base, using ontologies and building a system of "semantics scrapers" that will access the semantic markups, integrate them appropriately into the growing knowledge base, and set up the material for the scrutiny of an editorial process.

To view full article, please click here.

[Post New Comment]
Mind·X Discussion About This Article:


	Feigenbaum's Challenge posted on 07/15/2003 8:24 AM by Kurzweillian	[Top] [Mind·X] [Reply to this post]
I agree wholeheartedly with Feigenbaum that the WWW is "an apple to be plucked" for data to support a computational intelligence. In fact, I believe this will be the primary contributing factor to the singularity. Quantification of data is definitely available, it is the qualification of this data that may prove to be a sticky point which expert systems will have to overcome. I feel however, that the Turing test overlooks one important consideration. When machines become "spiritual" their answers will be so correct, inciteful and so eloquent that no human answer could compare. Stand back and watch it happen.


	Re: Feigenbaum's Challenge posted on 07/15/2003 7:31 PM by PsyTek	[Top] [Mind·X] [Reply to this post]
Kurzweillian wrote: When machines become "spiritual" their answers will be so correct, inciteful and so eloquent that no human answer could compare. And their answers will be commensurately unrecognizably so by human standards, so would we think it had become spiritual, or simply nonsensical? Cheers! PsyTek.


	Re: Some Challenges And Grand Challenges For Computational Intelligence posted on 07/15/2003 1:17 PM by clarkd	[Top] [Mind·X] [Reply to this post]
The idea of using the Web in it's current form for gathering correct data for a knowledge database won't happen. The web is 98% garbage and the last 2% has no measure of it's correctness. Using data that is so flawed will cause many more errors and “dead ends” than any good that could be had. If certain web sites were found that did have good quality data, then that data could be used but finding those sites would have to be the work of a human not a program. The idea that it will be the data and not so much the algorithms that manipulate that data that will be the hardest to come by, I think is very astute. If only there was a database that was in a form that any researchers could access (for free) and add to, that people could experiment with! Good ideas in the article.


	Re: Some Challenges And Grand Challenges For Computational Intelligence posted on 07/15/2003 2:37 PM by /:setAI	[Top] [Mind·X] [Reply to this post]
The web is 98% garbage so are all information pools


	Re: Some Challenges And Grand Challenges For Computational Intelligence posted on 07/15/2003 7:42 PM by PsyTek	[Top] [Mind·X] [Reply to this post]
Hi /:setAI, Yes, I tend to agree. In any case, "accuracy of information" I find to be worth far less than the ability to discern agreement or conflict within a body of "knowledge". Humans have the power to be poetic, express and appreciate nuance, precisely as we do NOT hold every word to be fixed in its meaning, but rather to be mitigated by context. Likewise, our ability to be resourceful, creative and "appropriate" comes from out holding all prior "knowledge" suspect and subject to revision or contextual mitigation. People who think that intelligence is the gathering of countless "facts", like pieces of a giant jigsaw puzzle, will be very disappointed to find it is impossible to fit those pieces together, or many different ways are possible. Even though 98% of what is "out there", if taken at face value, is garbage, a system that can seek "alignments" and relations among that morass may be able to discern more truths from it than it presents on the surface. But they will not be obvious, and simply piling on more "right data" will not help. Understanding comes from the ability to perceive "fitness", and in our complex world, that fitness is going to have to be nimble and dynamic. Cheers! PsyTek.


	Re: Some Challenges And Grand Challenges For Computational Intelligence posted on 07/16/2003 10:12 PM by claireatcthisspace	[Top] [Mind·X] [Reply to this post]
Tony, Very clever. Claire


	Re: Some Challenges And Grand Challenges For Computational Intelligence posted on 09/13/2003 2:46 AM by jontait	[Top] [Mind·X] [Reply to this post]
I think that most people are going about this the wrong way. Information is contrast: a way to consider data relative to other data. If the only place in the world you live in is the town you were born in, it doesn't matter how well inside and out you know that town, you still can't compare it to living in other towns. Additionally, I would like to address the web containing junk issue. If you look at each webpage and try to learn something, of course you will only find something useful < 15% of the time. But step back and take a look at the bigger picture. The web is a fantastic way to statistically observe contextual contrasts. Cross-referencing this with a dictionary, encyclopedia, thesaurus, etc. has boundless potential. I recently did a relatively simple school project to prototype this idea and the results were very promising.


	Re: Some Challenges And Grand Challenges For Computational Intelligence posted on 07/20/2003 5:00 AM by Thomas Kristan	[Top] [Mind·X] [Reply to this post]
The web is 98% garbage and the last 2% has no measure of it's correctness. Google with "site:.edu" - and the percentage of garbage is bellow 20. If not 10. - Thomas


	Re: Some Challenges And Grand Challenges For Computational Intelligence posted on 07/20/2003 12:48 PM by Mentifex	[Top] [Mind·X] [Reply to this post]
http://www.ai-forum.org/topic.asp?forum_id=1&topic _id=8343 -- AI has been solved

Brain Archive