This is a fascinating game! It seems that if you pick something fairly generic: a jet, a canyon, a dog, it doesn't have much trouble. It tends to bog down with something really specific: a red-tailed hawk, an enchilada. What's truly interesting to me is that it obviously uses a much different strategy than a human. I've played about 10 times now, and I think I have an idea of what it's doing--it's got a finite list of questions, and then it simply catalogs the "correct" answers for each term.

It seems to select the next question based on the one that splits the remaining possible terms in half--which would account for the seemingly random questions.

But it must do much more than that, because it needs to account for the possibility that the user responded "incorrectly."

I noticed this morning when I played, it gave me more options--more possible answers. I wonder if the program has been modified since yesterday, or if it adjusts the possible answers based on server loads. I also wonder if it somehow "learns" as more people play the game.

I also found this quote from the home page to be quite interesting: the game does not take anonymous players as seriously as members. I've always wanted a computer to take me seriously! I think I'll sign up!

Dave

i think that the program is self-adjusting and/or being tweaked by the designer, and that someone is using it for developing either a knowledge base off the answers, or for refining the inference engine for determining the answers, as part of a larger ai project.

This game is incredible...It guessed mine--sword, without any difficulty.