AI that learns from RAM changes

http://www.youtube.com/watch?v=xOCurBYI_gY

Think Markov Chains building from the temporal RAM states. NES (and robot microcontollers too!) have a little memory, so you can pattern-match it directly.

And you can apply the idea to self diagnose various software! I.e. some unusual patterns would be a sign of errors.

The problem is that he bruteforces the search. Hidden Markov model could have been used instead of emulation. I guess evaluating a markov chain is much more efficient than running an emulator. Moreover, markov chain can easily span minutes, while emulator evaluates just a few frames forward.