Bringing up Baby Robots

By Edward Cone  |  Posted Friday, September 17, 2010 18:09 PM

By Tim Moran

The quest by researchers to create robots that are as smart as -- or smarter than -- humans continues apace. We're not talking here about the recent wave of telepresence bots, such a Vgo and Telenoid; these are mere shells designed to mimic and interpret the speech and movement of real humans that are wirelessly attached at the other end.

No, this is about AI, serious artificial intelligence as pursued by the same fine folks who brought you the internet -- the Defense Advanced Research Projects Agency, aka DARPA. Writes Wired's Katie Drummond: "A Pentagon-funded scientist has come up with a comprehensive program to turn today's robots into tomorrow's A.I. overlords. Step one: Imbue them with toddler-level intelligence. Step two: Run them through a 'cognitive decathlon' of tests. And, finally, use programmed learning abilities and human instruction to turn bot tots into supersmart A.I. agents 'that [can] learn and be taught like a human.'"

Shane Mueller, a researcher and cognition expert with Applied Research Associates, wants to start out with a robot that has the mental and physical capabilities of a two-year-old and, through training and experience, have the robot develop through stages. Mueller explained that: "There were many motivations for this target, but one central notion is that if one could design a system with the capabilities of a two-year-old, it might be possible to essentially grow a three-year-old, given realistic experiences in a simulated environment."

Mueller devised a testing schema that measures things like visual recognition, search abilities, knowledge learning, concept learning, and motor control. From that tot-level baseline, Mueller expects that the "AI agents," as he calls them, would "gradually learn from both surroundings and an instructor, advancing to more advanced cognitive capacities."

In the end, Mueller is out to create an AI robot that mimics human thinking and behavior, warts and all: "When consistent errors or imprecisions are found in human performance, it may go hand-in-hand with a flexibility and robustness, such that if our performance were better, it would also be more brittle."

Frankly, I really have no idea what that means. Perhaps that's because I, like the tot bots, have the mental capabilities of a two-year-old. Actually, as the father of three, I am much more interested in how Mueller would react to his tot bot saying, "Daddy? Daddy? Daddy?" 30 or 40 times in a row and asking for yet another battery charge in the middle of the night. Brittle, indeed.