Alex Waibel is disappointed in his computer. “It doesn’t care what I do and who I am and where I sit,” says the director of Carnegie Mellon University’s Interactive Systems Laboratories. “It just doesn’t do anything until I do something-until I hit some button.” Splitting his time between his lab at Carnegie Mellon and a sister lab at Universitt Karlsruhe in Germany, Waibel is working to free humans from that forced interaction with machines. His model for the ideal computer? “A good butler or a good secretary-someone who invisibly hovers in the background, guesses your very needs, and serves them up before you even ask.” That way, he says, humans would be free to interact with other humans and, as he puts it, “do the human thing.” Computers would observe what the humans were doing-and understand it sufficiently to guess how to help the humans out. Sitting at a table in his office with Technology Review senior editor Rebecca Zacks, Waibel explains how that might play out. “I want to talk with you and then,” he says, craning his head toward his desktop machine, “say, Oh, by the way, write a letter to so and so and tell him I can’t do the review.’ But how should the machine know that I’m now talking to it and not to you? If I say something about deleting the files, I don’t want to have my computer go off and delete all the files. It needs to know who is being addressed.” That’s just one of a number of obstacles standing between Waibel and the computer of his dreams. He told Zacks about a few more hurdles and showed her what his team is doing to clear them.