BRETT working through a given task (via UC Berkeley)
These days artificial intelligence can do some amazing things from drawing pictures to removing stalled cars, but it knows how to do these things through explicit instructions. Is it possible for a robot to learn a task on its own? A research team from the University of California, Berkley seems to think so. BRETT (Berkley Robot for the Elimination of Tedious Tasks), a Willow Garage Personal Robot 2 (PR2), has the ability to complete tasks on its own through trial and error, much like humans. It can assemble a toy airplane or Lego bricks by using neural network-based deep learning algorithms to master certain tasks.
This technique is loosely inspired by the neural circuitry of the human brain when it interacts with the world. In turn, it helps the robot recognize patterns and categories its receiving. Unlike other forms of artificial intelligence, you rarely have to tell the robot what to do via new code – just give it a task and enough time to figure things out. There's even a reward system that scores BRETT on how well it learns a new task. The movements and strategies that allow it to finish the task are scored higher. The information is then relayed across thousands of parameters in the neural net.
While it would be amazing to have a robot assemble your Ikea furniture it's not exactly ready for the real world. It takes quite a bit of time for BRETT to complete a task; ten minutes when told exactly where to start and stop and three hours if its learning things by itself. With further development the team hopes the technology will improve to handle more data over the next several years.
“With more data, you can start learning more complex things,” said Professor Pieter Abbeel of UC Berkley's Department of Electrical Engineering and Computer Sciences.“We still have a long way to go before our robots can learn to clean a house or sort laundry, but our initial results indicate that these kinds of deep learning techniques can have a transformative effect in terms of enabling robots to learn complex tasks entirely from scratch. In the next five to 10 years, we may see significant advances in robot learning capabilities through this line of work.”
BRETT is part of a new People and Robots Initiative at UC's Center for Information Technology and Research in the Interest of Society (CITRIS). The goal of this new campus, multidisciplinary research initiative is to keep advances in artificial intelligence, robotics, and automation aligned to human needs. The team's research is supported by The Defense Advanced Research Projects Agency, Office of Naval Research, U.S. Army Research Laboratory, and National Science Foundation.
C
See more news at: