I remember seing the exact same thing hosted at MIT a couple of years back. The term AI here is a bit of a misnomer since its intelligence is actually driven by the users of the tool and not the tool itself. I think this tends to be the first project of most graduate students studying AI. If it makes a guess, and guesses wrong - then it asks you to tell it what the answer is as well as a question that would lead to the answer, and every time thereafter it'll know the correct guess for that specific boolean tree path you've navigated it through. So, the tool doesn't work that well for the first 1000 times or so that people use it but as time goes on it just gets better and better at guessing the right answer. Quite effective!

- trs