To extrapolate is human: unless we've overlooked to the "PUSH" sign and are frantically pulling on a door, we easily adapt to differing styles of handles and knobs, figuring out intuitively whether to slide, rotate, push down, and so on. It's not so easy for robots, and if we eventually want to have robots that interact with humans in natural environments that aren't completely specified and controlled, we need them to develop a better understanding about how the world works.
Researchers at Google have been developing learning strategies that can be distributed across a group of robots, and MIT Technology Review interviewed Assistant Professor Stefanie Tellex of Brown University's Department of Computer Science (Brown CS) for her thoughts on the project.
You can read the full article here.
For more information, please click the link that follows to contact Brown CS Communication Outreach Specialist Jesse C. Polhemus.