DARPA’s Robot AI Learns To Cook By Watching YouTube Videos

The Defense Advanced Research Projects Agency (DARPA), which commissions advanced research on behalf of the Department of Defense (DoD), published a press release on January 29, 2015, indicating a DARPA-funded research program at the University of Maryland has developed a system which enables robots to process visual data from a series of YouTube videos on how to cook.

Reza Ghanadan, DARPA’s Defense Sciences Offices program manager, said that DARPA’s Mathematics of Sensing, Exploitation, and Execution (MSEE) program, which funded the research at the University of Maryland “focused on sensing” and understanding what’s happening in the visual scene instead of merely recognizing and identifying objects.

“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.”

She went on to say that the system developed “allows robots to continuously build on previous learning,” and that this could have a significant impact on both teaching as well as training.

“This system allows robots to continuously build on previous learning—such as types of objects and grasps associated with them—which could have a huge impact on teaching and training (…) 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.”

DARPA’s research was presented at a meeting of the Association for the Advanced of Artificial Intelligence on January 29, 2015.

The UMD paper was authored by the university’s Yezhou Yang, Cornelia Fermuller, Yiannis Aloimonos, and NICTA’s Yi Li, from Australia.

An NBC News report summarized the school’s recent achievement as a system which “translates basic actions seen in YouTube cooking videos to real-life actions carried out by a Baxter humanoid robot.”

UPI report that what’s revelatory about this new development is the fact that robots are now “being programmed to continuously build on prior information and learning.”

What do you think billionaire tech entrepreneurs Bill Gates and Elon Musk think about the university’s recent advances in artificial intelligence in light of their recent commentary exhibiting a fear of robots turning on humans in a perhaps not so far fetched Terminator-esque future?

[Image via MSEE/University of Maryland]