I think I heard this story a while ago, but now is the first time I actually remembered the address. An electronics engineer made a small device that automatically plays the game Guitar Hero, scoring better than any human ever possibly could, because it never makes mistakes.
Guitar hero is a "rhythym game," in which players use a plastic replica of an electric guitar to press buttons as they appear on a scrolling musical scale, in tribute to Bach's famous quote about musicianship: “I just press the right keys (buttons) at the right time and the organ does the rest. ” The guitar vaugely resembles the actual action of playing a guitar, in which a guitarist holds down some of the strings to change the effective length, and thus the pitch, of the vibrating string. Players of Guitar hero have about two seconds head-warning before they need to press the respective button.
Anyway, the machine receives a video signal, analyzes this signal, and uses it to determine when to send the button-push signal back to the game console. Two seconds is enough time for the computer to have completed its analysis, so the machine literally can't fail.
Why do this project? For one, it's an interesting look at video-analysis. Visual recognition is currently one of the weaker areas in computers. Show a computer a picture and it will interpret it only as a matrix of colors. Attempts to recognize pictures of people, useful for "We have a picture of a person walking into an airport. Is this person one of these people who are wanted criminals?" have been foiled by wearing different glasses, growing or shaving facial hair, or other things that wouldn't fool a human for even a second. There is big money, then, in getting computers to actually understand what it is that they are looking at.
For another, it's the "because I can" effect. Getting a computer to exactly copy a human action is an impressive boast.