Facebook wants to teach robots how to walk

Facebook wants to teach robots how to walk

Artificial intelligence researchers on Facebook have revealed details of their robot training efforts, including an initiative to teach a six-legged robot how to walk. The company said its devices are smarter and much faster.

Although Facebook does not sell robots, researchers use many of them. Facebook, as a global social networking platform, relies heavily on artificial intelligence and automated learning systems to keep harmful content away from its platform.

The social networking company announced three robotic projects that hope to contribute to solving the ongoing challenge of building artificial intelligence systems that do not have to rely on large amounts of data to learn new information.

These projects are tracked by the company's Artificial Intelligence Department, a division that works independently of Facebook's popular suite of applications and services.

The company says its research in the field of robotics and artificial intelligence is used to promote artificial intelligence at the industrial and academic level, and that computer scientists are not conducting research for the purpose of integrating technology into its products destined for users.

To reach that goal, Facebook is conducting research to teach robots how to recognize the world, like the human way, in collaboration with computer scientists from New York University.

"Our goal is to reduce the number of interactions the robot needs to learn to walk, so that it takes hours instead of days or weeks, even if the device does not have information about its current environment," the researchers write in a publication.

They designed techniques, which include improved Bayesian statistical inference, as well as enhanced model-based learning, to be deployed to work with a variety of robots and environments.

Automated systems education is a long-term challenge that Boston Dynamics faced with its SpotMini robot. Google also faced its DeepMind artificial intelligence system when it trained in 2017 as a model for moving toward more flexible systems.

The Facebook method for teaching robots how to walk involves putting sensors in the joints of each leg of the robot stalks and using self-enhanced tutoring as a means of training artificial intelligence through repetitive simulations that do not require task-specific training data.

Google's artificial intelligence researchers in April published a detailed study of the efforts to teach a mechanical arm how to be curious about the physical world. Curiosity was used as an incentive that could help robots learn more quickly.

A month ago they developed a method that allows robots to mimic the sense of touch in order to expand their ability to move things and manipulate them beyond the realm of computer vision, without giving any specific training data.

The largest Silicon Valley companies, such as Facebook, are investing heavily in AI. Market research firm Gartner predicts that the value of artificial intelligence will reach $ 1.2 trillion in 2018, a 70 percent increase over 2017.