A robot may capture video of itself to review how it well it did, then analyze and build on that information as it keeps improving over time. Fanuc’s machines will feed what they learn into a neural network that other robots can learn from and contribute to as well, reported MIT Technology Review.

Industrial robots are big and dangerous and really good at doing a single task over and over again with exacting precision.

But when the production flow changes, it can take days for an engineer to write a new teaching program and get all the massive machinery onboard. With reinforcement learning through AI, however, robots on an assembly line can teach themselves to take on a new task overnight.

The idea is that when Fanuc’s robots learn together, they learn faster. And Nvidia’s technology is especially good at parallel processing, meaning it can simultaneously handle thousands of computational tasks.