Facebook is surely a high-tech firm, but it is not one you might essentially associate with robots. On the other hand, as the company disclosed this week, that is exactly where its scientists are seeing next—attempting to see how tests in robotics can boost its work in AI.
This is not strange for big tech firms. A lot of companies, comprising Nvidia, Google, and Amazon, employ robots as a platform to discover opportunities of AI research. Controlling bots is, in many manners, harder than challenges such as playing video games and board games. With these latter tasks, scientists have found a way to simulated game surroundings, which permits AI agents to learn and play at accelerated paces. There is no such shortcut for skilling robots.
“The best thing about robotics is that it occurs in the real world and in real time,” claimed co-managing director of Facebook’s AI research labs, Antoine Bordes, to the media in an interview.
The study is wide-ranging, and the firm has shared info about a trio of papers. The first comprises getting a 6-legged machine to skill itself how to walk via trial & error, the second is about using the “curiosity” to assist bots learn quicker, and the 3rd is about employing a sense of touch to assist a robot attain simple jobs such as playing with a ball.
Speaking of AI, when encountered with difficult questions about how Facebook will eliminate terrorist material from its service, CEO Mark Zuckerberg provides an easy answer: AI will do it. But as per Yann LeCun, chief AI scientist at Facebook, AI is years away from being capable of fully shouldering the burden of restraint, specifically when it comes to managing live video.
Talking at Facebook’s AI Research Lab at an event in Paris previous week, LeCun claimed that Facebook was years away from employing AI to regulate live video at scale, claims media.