@thrill: Lots of people can tell you at a high level how AI will work. It is a matter of getting people together to hash out the details. This is a lot closer to Wikipedia than it first appears.

@Richard W: Liability is a problem, but not a technical one. Let’s build it and then figure out the legal issues. It is actually not that hard to build something much more reliable than humans.

@Will: Some of those words about chess engines are 4 years old now. Thanks for the new info, I might refresh them. I completely disagree that a proprietary license (which is different from the profit motive) is necessary to write good software. In fact, I have found many cases where proprietary software is technically inferior to its counterpart in free software, just like how Wikipedia is better in many ways than its alternatives. My book discusses a number of them.

@Mike S: We might build transponders in every street corner and such, but in the meanwhile, we need cars that have vision systems. And once we’ve done that, we can just teach it to read Stop signs. There is OCR code out there, but those people aren’t working with the driverless people.

Beyond the minor point that Deep Junior is not even in the top ten of modern chess programs (http://ssdf.bosjo.net/), the best program is currently Rybka, which is closed-source commercial. However, there are open-source collaborative efforts, like Stockfish, which are making a run at the top spot. Current chess theory, you are correct, has nothing to do with AI, so it’s not really relevant to the discussion. I think the profit motive will always drive the best R&D, and open-source collaboration is useful in the absence of profit.

We all know how an encyclopedia will “work” — we yet don’t know how AI will work, and that’s the value of having multiple teams working on the problem — we need to yet find *the way*, and that involves multiple explorations.