Teaching in Y2K

The Y2K glitch is highly inspiring. The central message of millenial doomsters is humorously overblown, but a peripheral truth -- that most aspects of life are now being run by machines -- is worth paying attention to. And what makes my job so rewarding is the knowledge that the generation of students I'm teaching is the one that will make these machines intelligent.

Although no-one is about to build Star Trek's Lt. Commander Data, artificially intelligent (AI) decision making by computers is already important and practical (AI technologies such as automated discovery, reinforcement learning, constraint satisfaction and even game-playing now have payoff opportunities over 1 and 2 year periods, fermenting many startup company opportunities).

But the icing on the cake is that thinking about the basics of rational decision making involves elegance and cuteness. Many parts of AI are stories that pretty much tell themselves without the instructor's help. We often begin with informal notions of occasions where we humans need to be intelligent in our lives ("I know that when I'm negotiating for a car I shouldn't reveal my price too soon, but what SHOULD I do?", "I'm feeling sick: how can I make myself confident that I don't have the Ebola virus?"). We continue by turning this into mathematics and data structures. We then gasp with horror at the apparent computational intractability of doing the optimal thing. Then we sometimes are able to heave sighs of relief when there are ideas to be stolen from previous researchers in disciplines such as AI, algorithms or statistics that save us. And then we think about the computers that are today managing information in, say, Walmart, Blue Cross or your car's engine. The story frequently concludes with a realization that we could make our lives safer, more comfortable, less resource intensive and (most important) more Star-Trekky, by actually going out and deploying what we've discussed. Y2K is an exciting time to be teaching AI.