In one of Iain M. Banks' Culture novels — I think it was Consider Phlebas — one of the vast A.I. Minds laments to its human ward that even with all its vast computing power it can only calculate the scattering of billiard balls with any accuracy for less than a few moves across the table. While that may hold true in the real world for the moment, the "Deep Green" robotic pool playing system is already dropping cellulose spheres at a "better-than-amateur level". Granted, its matrix of gantries and cameras isn't predicting the moves of every ball from games' beginning to end, but its ability to adaptively analyze a table is still impressive.
Plus who doesn't love a robot with something labelled "pneumatic break cue actuator"?
Computer.org did a nice profile of the system in January, including lots of details from Canadian scientists on the team explaining how they built Deep Green with mostly off-the-shelf parts.
Why build a pool-playing robot at all? Just a few moments listening to the nearly monotone female voiceover while the video shows a whipping, articulated white robot leering from the ceiling makes it clear: this is where GLaDOS will be born.