Friday, March 20, 2009

Sad News for Lazy People


In LiveScience’s “Robot Madness”, Jeremy Hsu points out the differences in true artificial intelligence and robotic AI (ie, formal logic). For many people, this particular issue may crush their hopes of having robot butlers that do everything for them. Currently, there are robots that do pretty cool things, like sort through thousands of pages of information quickly or play chess, but trying to program and teach one to fetch you something from the refrigerator is a completely different story. Hsu states that the fundamental problem in creating a “robot Jeeves” is connecting abstract logic with real-world meanings like “beer” and “fridge handle”. Some of the things that robots need to be able to learn to complete these types of tasks (don’t crush the can, differentiate between cans in the fridge, don’t drop the drink into someone’s lap, etc) can only be done by replicating the human perception that we use to learn. Obviously, that would take painfully complicated logic, and even then would not be as useful as having an actual person complete the task. Another problem that scientists have encountered with robots is their inability to see the world as well as humans (colors, shadows, objects): another hindrance that prevents the creation of a completely independent robot. Robotics expert Chad Jenkins from Brown University has excelled in this field with his robots that read and obey non-verbal commands. Using hand gestures, Jenkins’ team coded a bomb-disposal PackBot (normally controlled by a human soldier)to obey gesture commands to halt, wait, follow, and door breach. In the future, Jenkins hopes to create a PackBot that could be “trained” by a soldier to follow gestures and remember them. Although this sort of research can easily lead to one person being able to control a team of robots to perform specific tasks, researchers still have quite a way to go before robots can do very human tasks without any human supervision or guidance.

0 Comments:

Post a Comment

<< Home