Robots? Learning from YouTube videos? We sure hope they don't check out any of the thousands of (possibly fake) prank videos on the giant video site; or, worse, that Chocolate Rain guy. The last thing we need are robots trying to sing in a rich, quasi-pop bass.
Researchers at the University of Maryland are currently working to train robots to act on visual stimuli, and their research has achieved some interesting early results. The project, funded by DARPA's Mathematics of Sensing, Exploitation, and Execution program, initially concentrated on the most important robot task of them all—teaching machines how to cook.
No, a robot isn't about to start telling off its human patrons all Gordon Ramsay-style. In the project, researchers "trained" the robots to react to objects in their environment by giving them YouTube cooking videos to watch. The robots processed the objects in the video, figured out what they were, and then applied that "knowledge" to objects in a real-world environment. The robots correctly identified and grabbed specific objects and, more interestingly, manipulated them the same way they saw others manipulating kitchen objects in the YouTube videos—all without any additional input or extra programming.
"The MSEE program initially focused on sensing, which involves perception and understanding of what's happening in a visual scene, not simply recognizing and identifying objects. We've now taken the next step to execution, where a robot processes visual cues through a manipulation action-grammar module and translates them into actions," Reza Ghanadan, a program manager at DARPA, said in a statement.
Researchers have also been investigating new ways for the robots to retain this knowledge and "teach" other robots what they know. These two steps could greatly shorten the amount of time it takes to get new robots online and cooking away. Or, at least, grabbing objects around the kitchen (or wherever).
"Instead of the long and expensive process of programming code to teach robots to do tasks, this research opens the potential for robots to learn much faster, at much lower cost and, to the extent they are authorized to do so, share that knowledge with other robots. This learning-based approach is a significant step towards developing technologies that could have benefits in areas such as military repair and logistics," Ghanadan said.
At last year's SXSW, IBM's Watson used his smarts to algorithmically concoct recipes, and attendees voted on Twitter via the #IBMFoodTruck hashtag for the daily dishes they most wanted Watson to cook up.
Later that year, IBM and Bon Appetit teamed up for a pilot app called "Chef Watson with Bon Appétit," which used the power of Watson to help chefs discover new recipes. In April, those culinary smarts will be available as a cookbook.
Post a Comment