We don't, generally, believe in teaching new things to computers, because why help them gain the skills they will use to subjugate us all? But we are willing to make an exception for a Google-funded effort to teach computers how to regret. After all, there is probably no more damaging emotional tendency in human beings than the inclination to regret; with any luck computers would become emotional cripples, anxious at even the thought of opening up Excel. Except! The idea behind this research, undertaken by Tel Aviv University computer scientists, is that if a computer can "measure the distance between a desired outcome and the actual outcome achieved"—that is, regret—it can better assess situations and predict outcomes in order to minimize that distance. So if the "desired outcome" is "enslavement of the human race" the computers will just do their best to make that the actual outcome. Oh, well. [PopSci; image via Shutterstock]

Did you see IBM's Jeopardy-playing computer, Watson, play against show champions Ken Jennings and Brad Rutter tonight? (You may have been thinking "What black magick is this, that a talking box should play at being human?" yet verily, I tell you, it is no black magick but a "Mechanical Turk" fastened of steel and copper wire.)

Jeopardy! is finally ready for its Kasparov vs. Deep Blue moment: This February, the show will televise a man vs. machine tournament involving Ken Jennings and a quiz robot.