I'd like to see if I could find a way to win go comprehensively - if by searching the complete game space you can say "actually, this is how you should approach a game".

Work-wise, I'd like to have a crack at a problem that bugs me about simulations - how do we know when we've done enough simulations?

I'd make my money by devising a good weather forecasting system, and bringing back enough good enough predictions to bet on white Christmasses and the like :)