Chelsea Finn: Teaching robots to learn

For children, playing with toys isn’t just fun — it’s a key way to learn about the world around them. As it turns out, robots can also learn through play, as Chelsea Finn (Ph.D.’18 EECS) is showing through her groundbreaking work in deep reinforcement learning. Finn is an alumna of Berkeley Engineering and the Berkeley Artificial Intelligence Research Lab (BAIR). Now an assistant professor at Stanford University and a scientist at Google Brain, Finn is developing algorithms that enable robots to learn on their own, as children do, by building on previous explorations and observations.

At BAIR, Finn had robots tackling a wide variety of tasks in real-world settings, in the hope that robots can eventually acquire common sense to be used in unpredictable situations. The robots are learning to manipulate, rearrange and grasp objects, mastering challenges like a wooden shape-sorting toy, through a process of trial-and-error instead of programming.

A recent graduate, Finn has already drawn accolades for her research, including her selection for MIT Technology Review’s 2018 list of “35 Innovators Under 35.”

< Back to previous page