Is it really "making better decisions than a human using a pretty reasonable approximation of the same mechanism as a human" though? To a layman, it looks rather more like an example of a high-powered calculator estimating Nash equilibria by crunching vast amounts of data. More human-like than Deep Blue, certainly, because it's not able to evaluate all possible game states so there's a bit of pattern recognition going on to cut down its search space, but it's succeeding in Go because with Go the higher level of abstractions humans work at is a coping mechanism for not having enough calculating power to simulate outcomes, not the only way of simulating possible outcomes.

The issue that the real world is mostly not composed of optimisation problems with possible moves and win conditions as neatly and completely described in code and as replayable as Go is probably a thornier one for building AGI (as opposed to useful tools) through learning processes than the adequacy of the hardware...