Over at the Huffington Post, UC Berkeley roboticist/artist Ken Goldberg writes about the future of social robots that help each other learn to better navigate a world they never made. From HuffPo (image of Goldberg's TeleGarden):
Our robots are signing up for online learning. After decades of attempts to program robots to perform complex tasks like flying helicopters or surgical suturing, the new approach is based on observing and recording the motions of human experts as they perform these feats. Statistical Robot Learning, pioneered by researchers such as UC Berkeley's Pieter Abbeel, infers the underlying intentions of experts by analyzing patterns in their motions. My students and I are working with Abbeel on a new project to incorporate models of sensing so that these robots can cope with noise and adapt to changing conditions…
As humans embrace new forms of social media to keep connected with friends and colleagues, our robots are becoming increasingly sociable. Researchers at Google and several university labs are working on "Cloud Robotics," where robots benefit from four aspects of the Internet: (1) the availability of thousands of cloud-based processors to compute solutions remotely, rather than onboard the robot, (2) vast databases of information describing the physical properties of environments and commercially-available objects, (3) the ability of robots to share information with other robots about past successes (and failures), and (4) the availability, when all else fails, to contact remote human operators to ask for advice.
