By voting on Jyte's automatically generated claims, we will be teaching it to understand human language at a higher level.

you can't just take a data-driven approach. natural language understanding is required -- or at least the claims would have to conform to some syntax. if a claim is not well-formed, you won't be able to do any sort of inference on it.

You're preaching to the choir.

Obviously "understand" is an overstatement in the short run; I was in a hurry when I claimed this one.

Have you ever tried the program "MegaHal"? It uses a hidden markov model to generate text based on a corpus, which is built through interacting with it. A neat way to make text that almost makes sense.

No, is it better than Eliza?

"Syllogism based AI has been done to death" Has it been done with the help of a corrective social network?

Whether it's a true AI or not, there are some cool things we could do by making it machine readable.

Well jyte isn't just a knowledge base; it's reactive. It might not be the most accurate, but that wasn't the goal. Likewise, it might not be the ideal way to approach artificial language acquisition. What it has, though, is metadata on each claim giving context to varying degrees, and a notion of the reputations of the claimants. Eventually I wouldn't be surprised if a machine with access to the jyte dataset got pretty good at predicting who would vote which way on a given claim, or even guessing which user made a claim based on its "style."

I wonder if MegaHal could've come up with this one. Anyway, I'm a little bothered by the attitude that it's been "done to death." So has music, beer, automobiles, technology, writing, and virtually every other field of human endeavor.

If anything, AI, machine learning, and language acquisition are in their infancy compared to some other fields. Maybe they'll never work, but, as with anything else, if it worked the every time, it wouldn't be technology.

I guess I don't know what "syllogism-based" includes and excludes. Unless you're promoting emotional-based or intuition-based AI, I don't see any way to avoid building logical relationships between concepts. Further, I don't see a representation for a concept that doesn't involve a degree of language and logic. Is human intelligence syllogism-based?

Which claims are you talking about? Are you talking about a desired future feature, or is there a user entering in automatically generated claims, or what?

A future feature.