#108 – Sergey Levine: Robotics and Machine Learning

Sobre este episodio

<p>Sergey Levine is a professor at Berkeley and a world-class researcher in deep learning, reinforcement learning, robotics, and computer vision, including the development of algorithms for end-to-end training of neural network policies that combine perception and control, scalable algorithms for inverse reinforcement learning, and deep RL algorithms.</p> <p>Support this podcast by supporting these sponsors:<br /> &#8211; ExpressVPN: <a href="https://www.expressvpn.com/lexpod">https://www.expressvpn.com/lexpod</a><br /> &#8211; Cash App – use code &#8220;LexPodcast&#8221; and download:<br /> &#8211; Cash App (App Store): <a href="https://apple.co/2sPrUHe">https://apple.co/2sPrUHe</a><br /> &#8211; Cash App (Google Play): <a href="https://bit.ly/2MlvP5w">https://bit.ly/2MlvP5w</a></p> <p>If you would like to get more information about this podcast go to <a href="https://lexfridman.com/ai">https://lexfridman.com/ai</a> or connect with @lexfridman on <a href="https://twitter.com/lexfridman">Twitter</a>, <a href="https://www.linkedin.com/in/lexfridman/">LinkedIn</a>, <a href="https://www.facebook.com/lexfridman">Facebook</a>, <a href="https://medium.com/@lexfridman">Medium</a>, or <a href="https://www.youtube.com/lexfridman">YouTube</a> where you can watch the video versions of these conversations. If you enjoy the podcast, please rate it 5 stars on <a href="https://podcasts.apple.com/us/podcast/artificial-intelligence/id1434243584">Apple Podcasts</a>, follow on <a href="https://open.spotify.com/show/2MAi0BvDc6GTFvKFPXnkCL">Spotify</a>, or support it on <a href="https://www.patreon.com/lexfridman">Patreon</a>.</p> <p>Here&#8217;s the outline of the episode. On some podcast players you should be able to click the timestamp to jump to that time.</p> <p>OUTLINE:<br /> 00:00 &#8211; Introduction<br /> 03:05 &#8211; State-of-the-art robots vs humans<br /> 16:13 &#8211; Robotics may help us understand intelligence<br /> 22:49 &#8211; End-to-end learning in robotics<br /> 27:01 &#8211; Canonical problem in robotics<br /> 31:44 &#8211; Commonsense reasoning in robotics<br /> 34:41 &#8211; Can we solve robotics through learning?<br /> 44:55 &#8211; What is reinforcement learning?<br /> 1:06:36 &#8211; Tesla Autopilot<br /> 1:08:15 &#8211; Simulation in reinforcement learning<br /> 1:13:46 &#8211; Can we learn gravity from data?<br /> 1:16:03 &#8211; Self-play<br /> 1:17:39 &#8211; Reward functions<br /> 1:27:01 &#8211; Bitter lesson by Rich Sutton<br /> 1:32:13 &#8211; Advice for students interesting in AI<br /> 1:33:55 &#8211; Meaning of life</p>