Robots capable of playing Ping Pong with humans have been around for more than 20 years. In 1988 MIT published "A Robot Ping-Pong Player: Experiments in Real-Time Intelligent Control (AI)" by Russell L. Andersson. Ping Pong is interesting to roboticists because it requires understanding of dynamic environments, accurate real-time vision, fast actuation, and intelligence to play the game with a winning strategy. (Andersson went on to write an IEEE paper on an aggressive trajectory generator for robot ping pong.) Andersson's machine, built at Bell Labs, was made from a bulky industrial robot arm combined with multiple video cameras, used to reconstruct a 3D view of the playing space. Since then things have improved quite a bit. Andersson's project was recreated by Japanese researchers using a PC, two cameras, and a four axis SCARA arm with good results. Mechatronics students at the University of Adelaide also built a modern version of Andersson's Ping Pong robot. The most impressive robot ping pong player to date has to be Topio, a humanoid robot from Tosy Robotics that was able to beat many human players at the 2007 International Robotics Exhibition in Tokyo. See the YouTube video of Topio in action. While they don't actually play ping pong, automated ball throwers like the Newgy robots have become common training devices for ping pong players.