Facebook has previously shown interest in artificial intelligence (AI), but recently the firm lifted curtains from its new robotics research project where it aims to teach robots to learn themselves in order to advance AI.

The company published a blog post on Monday in which it detailed three of its ongoing robotics project. Every project focuses on looking for ways for robots to teach themselves from trial and error experience instead of relying on data prepared for training. The firm believes that this skill can have implications beyond robotics field.

“This work will lead to more capable robots,” Facebook wrote, “but more important, it will lead to AI that can learn more efficiently and better generalize to new applications.”

Facebook’s AI might be helping terrorists by removing evidence

Detailing about its projects, one of them is focused on enabling a six-legged robot to teach itself to walk. The tech firm is developing algorithms and sensors to reduce the amount of time it would take for the robot to learn to walk, even if it has no prior knowledge about its environment or physical abilities.

The second project centers on equipping robots with ‘curiosity’ to improve the learning process. Facebook found that using curiosity as a motivator could help robots learn more quickly, similar to the way humans learn. It applied this curiosity-driven technique to applications using a real robotic arm and also in simulations to explore and try new things.

Finally, the third project explores on enabling robots to learn through tactile sensing, or ‘touch’. The company created a method enabling robots to accomplish tasks by learning through touch without being given any particular training data.

Facebook is hoping that their robotics-oriented projects will lead to algorithms that can learn about the world similar to the way humans do, wrote Business Insider.

Roberto Calandra, a research scientist in Facebook’s AI division, told the publication. “The fact that we are trying to develop algorithms that work on real robots [will] help to create [AI] algorithms that, generally speaking, are going to be more reliable, more robust, and that are going to learn faster.”