[ad_1]
“Some bodies are already known to accelerate learning,” says Bongard. “This work demonstrates AI that can search for such organs.” Bongard’s lab has developed robot bodies adapted for specific tasks, such as giving callus-like coatings to the feet to reduce wear and tear. Bongard says Gupta and his colleagues expand on this idea. “They show that the right body can also accelerate changes in the robot’s brain.”
Ultimately, Gupta says, this technique could reverse our thinking of making physical robots. Instead of starting with a fixed body configuration and then training the robot to do a specific task, you can use DERL to allow the development of the optimal body plan for that task and then create it.
Gupta’s animals are part of a broad shift in how researchers think about AI. Instead of training AIs on specific tasks like playing Go or analyzing a medical scan, researchers are starting to throw robots into virtual sandboxes. POET, OpenAI’s virtual hide-and-seek arena, and DeepMind’s virtual playground XLand— and learning how to solve multiple tasks in ever-changing, open-ended training dojos. Rather than mastering a single challenge, AIs trained this way learn general skills.
For Gupta, free-form exploration will be key to the next generation of AIs. “We really need open-ended environments to build smart agents,” he says.
[ad_2]
Source link