Sign In

Communications of the ACM

ACM TechNews

Way Beyond AlphaZero: Berkeley, Google Work Shows Robotics May Be Deepest Machine Learning of All


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
The reserachers said having robots perform real-world tasks, as this one appears to do, presents the greatest challenge and the greatest opportunities, for reinforcement learning.

Researchers at the University of California at Berkeley and Google have summarized several years of experiments with robots using what's known as reinforcement learning.

Credit: Shutterstock.com

Researchers at the University of California, Berkeley and Google have released a report on lessons learned from years of experiments in training robots with deep reinforcement learning (RL).

The researchers found that robotics does not abide by the most basic assumptions in the RL paradigm.

They said latency in robotics "violates the most fundamental assumption of [the Markov Decision Process], and thus can cause failure to some RL algorithms."

The researchers also detail the real-world challenges associated with goals and rewards.

Said researcher Sergey Levine, "In a game like chess or Go, the RL policy will only be as good as the 'simulator' that it inhabits," but in the real world, "a robot can experience many of the same things that we experience ... and maybe even learn things that might surprise us."

From ZDNet
View Full Article

 

Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA


 

No entries found