A team led by Northwestern researchers is the first to use artificial intelligence to build robots from scratch, according to a Tuesday news release.
The unique algorithm designed a successfully walking robot in mere seconds that runs on a lightweight console compared to other AI systems that require larger storage and datasets.
The study published Tuesday in the scientific journal, Proceedings of the National Academy of Sciences of the United States of America, states this technology has the potential to carry out near-instantaneous design, development and debut of bots for medical, environmental, vehicular and space-based tasks.
McCormick Prof. Sam Kriegman, who headed the experiment, said the AI system has compressed billions of years of evolution and bypasses the bias of robotic designers.
“Evolution has no foresight,” Kriegman said in a Tuesday YouTube video. “It cannot see ahead of time if a mutation will be beneficial or catastrophic. What we discovered is a way to remove the blindfold and thereby compress billions of years of evolution into an instant.”
After the AI receives instructions, it works to add or remove particles to create the shape of the body and assess itself after each iteration to eventually achieve its goal. The robot’s locomotion is a result of pumping air into the body’s air muscles.
The entire design process — from a shapeless mold with no visible movement to a moving robot — took 26 seconds.
“I think when some people look at this robot they see a useless gadget. I see the birth of a brand new organism, and it’s a thrill to witness its genesis,” Kriegman said in the video.
Email: [email protected]
Twitter: @HabashySam
Related Stories:
— Northwestern researchers develop smallest-ever walking robot in crab shape
— Northwestern Robotics Club talks fighting robots, gears up for future competitions
— Northwestern students’ newest friend: AI program ‘Ask a Wildcat’ connects students and alumni
