Facebook is unveiling AI research that is working on touchy-feely robots that can learn tasks through trial and error, reports say.
"Much of our work in robotics is focused on self-supervised learning, in which systems learn directly from raw data so they can adapt to new tasks and new circumstances," a team of researchers from Facebook AI Research wrote in a blog post," Engadget reported.
"In robotics, we're advancing techniques such as model-based reinforcement learning . . . to enable robots to teach themselves through trial and error using direct input from sensors."
According to Engadget, the task they gave a six-legged robot was to teach itself to walk without any outside assistance.
Using a reinforcement-learning algorithm, the robot figures out a controller that helps it achieve forward locomotion, the tech outlet explained, reporting "the more experience the robot gains, the better it performs."
The researchers' next step was to induce curiosity.
"What we wanted to try out is to instill this notion of curiosity," Franziska Meier, an AI research scientist at Facebook, told Wired, noting that is exactly how humans learn to manipulate objects.
Finally, Facebook is working on teaching robots how to feel physically — with a "predictive deep-learning model originally designed for video," Engadget reported.
"Our work shows that such policies may be learned entirely without rewards, through diverse unsupervised exploratory interactions with the environment," the researchers concluded, Engadget reported.
According to the news outlet, the Facebook researchers got a robot to successfully manipulate a joystick, roll a ball, and identify the correct face of a 20-sided die.
© 2025 Newsmax. All rights reserved.