menu Home chevron_right
NEWS & CULTURE

Sergey Levine: Robotics and Machine Learning | Lex Fridman Podcast #108

Lex Fridman | January 27, 2026



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.

Support this podcast by signing up with these sponsors:
– ExpressVPN at https://www.expressvpn.com/lexpod
– Cash App – use code “LexPodcast” and download:
– Cash App (App Store): https://apple.co/2sPrUHe
– Cash App (Google Play): https://bit.ly/2MlvP5w

EPISODE LINKS:
Sergey’s Twitter: https://twitter.com/svlevine
Sergey’s Website: http://rail.eecs.berkeley.edu/
Sergey’s Papers: https://scholar.google.com/citations?user=8R35rCwAAAAJ

PODCAST INFO:
Podcast website:
https://lexfridman.com/podcast
Apple Podcasts:
https://apple.co/2lwqZIr
Spotify:
https://spoti.fi/2nEwCF8
RSS:
https://lexfridman.com/feed/podcast/
Full episodes playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4
Clips playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41

OUTLINE:
0:00 – Introduction
3:05 – State-of-the-art robots vs humans
16:13 – Robotics may help us understand intelligence
22:49 – End-to-end learning in robotics
27:01 – Canonical problem in robotics
31:44 – Commonsense reasoning in robotics
34:41 – Can we solve robotics through learning?
44:55 – What is reinforcement learning?
1:06:36 – Tesla Autopilot
1:08:15 – Simulation in reinforcement learning
1:13:46 – Can we learn gravity from data?
1:16:03 – Self-play
1:17:39 – Reward functions
1:27:01 – Bitter lesson by Rich Sutton
1:32:13 – Advice for students interesting in AI
1:33:55 – Meaning of life

CONNECT:
– Subscribe to this YouTube channel
– Twitter: https://twitter.com/lexfridman
– LinkedIn: https://www.linkedin.com/in/lexfridman
– Facebook: https://www.facebook.com/LexFridmanPage
– Instagram: https://www.instagram.com/lexfridman
– Medium: https://medium.com/@lexfridman
– Support on Patreon: https://www.patreon.com/lexfridman

Tagged as , , , , , , , , , , , .

Written by Lex Fridman

Comments

This post currently has 41 comments.

  1. @lexfridman

    January 27, 2026 at 4:40 am

    I really enjoyed this conversation with Sergey. Here's the outline:
    0:00 – Introduction
    3:05 – State-of-the-art robots vs humans
    16:13 – Robotics may help us understand intelligence
    22:49 – End-to-end learning in robotics
    27:01 – Canonical problem in robotics
    31:44 – Commonsense reasoning in robotics
    34:41 – Can we solve robotics through learning?
    44:55 – What is reinforcement learning?
    1:06:36 – Tesla Autopilot
    1:08:15 – Simulation in reinforcement learning
    1:13:46 – Can we learn gravity from data?
    1:16:03 – Self-play
    1:17:39 – Reward functions
    1:27:01 – Bitter lesson by Rich Sutton
    1:32:13 – Advice for students interesting in AI
    1:33:55 – Meaning of life

  4. @jewishgenes

    January 27, 2026 at 4:40 am

    He said fetch the coffee for me hahaha. Nigga android but I toured for 15 years & played at a high level.

    It will be people like us helping robots understand androids like this guy 😂

  7. @NeuroReview

    January 27, 2026 at 4:40 am

    Rating: 7.6/10
    In Short: Robots are Key to Intelligence
    Notes: Sergey is a very well spoken computer scientist 'nerd' (meant in a good way!), but can speak at a normal level and explain things very well. He made some great analogies, especially connecting to Lex's glacier idea, and also his own about broken plates, to describe some complex ideas. He was clearly very thoughtful about Lex's questions (which were well done and less added on then other pods at this time). Something that really came out from this was Sergeys excitement to the idea of 'understanding intelligence', and how that is the main goal of his work studying robotics and machine learning. He gave off a very humble vibe, and was a really good guest. Would have loved a longer podcast, and feel that this is an underrated conversation.

  10. @TheArthurAbbott

    January 27, 2026 at 4:40 am

    Bro asks the meaning of life. Smirks.

    Life is full of meaning. How dare we be so reductionist to pick just one and hold onto it like the gospel truth. Is it to live? To love? To make an (positive) impact? To be happy? It doesn't boil down.

    But he smirks and asks the question, knowing the chance for a good response is well worth the ask.

  13. @skipperkongen

    January 27, 2026 at 4:40 am

    how legal would it be to build a robot, big or tiny, and let it roam freely outside where it could gather experience and have to survive on its own, e.g. keep power on its batteries? How legal would that be? Under what conditions would it be legal/illegal let alone ethical?

  14. @skipperkongen

    January 27, 2026 at 4:40 am

    Is simulation learning a thing? We start by learning in a simulator, then we use the policy in the real world and learn that it behaves different there compared to the simulator. That means there is an error in the simulator that we then make a step to fix and repeat?

  15. @skipperkongen

    January 27, 2026 at 4:40 am

    Do humans learn to wash the dishes by starting with an off-policy method of imitating our parents and then take it from there? Is it because breaking the dishes is associated with such a strong anti-reward, i.e. getting yelled at, that we burn that behaviour path to the ground and never use it again?

  17. @burkebaby

    January 27, 2026 at 4:40 am

    Sergey is one of the few Deep Learning researchers I have a huge appreciation. He is brilliant, prolific and an incredible communicator in one of the hardest areas of AI.

  19. @cogoid

    January 27, 2026 at 4:40 am

    Well put: 33:08 …the systems that we have now simply inhabit a different universe. [of pixels and sentences] And if we build AI systems that are forced to deal with … our universe maybe they will have to acquire our common sense…

  21. @TheAIEpiphany

    January 27, 2026 at 4:40 am

    I find it beautiful that Isaac Asimov had a profound impact on Sergey's passion for the field. It's funny how people have been influencing each other since the dawn of civilization.

  25. @vast634

    January 27, 2026 at 4:40 am

    I think the idea of using a human operator to steer the robot – only using the robots sensors and actuator commands – makes sense, to get a sense of the available feedback and control that the robot offers. And potential problems if those are not sufficient. Also to determine the concrete problem space an AI has to operate it. if a human operator cannot solve the problem, its unlikely that the robot could (apart from problems that require very quick reactions).

  30. @margrietoregan828

    January 27, 2026 at 4:40 am

    With the greatest of respect but both Lex & Sergey are here unknowingly & sadly irredeemably, conflating apples & oranges, as ‘computation’ (‘counting with digits’) & ‘thinking’ (‘intelligent use of information’) are two entirely different phenomena. Correct answers & solutions to robotics & separately to AI, will not be forthcoming until the all-important distinctions between these two domains are fully recognised & in which domains investigators, engineers & technicians comfortably & competently work.
    ‘Computation’ is nothing more than ‘counting’, for which ‘digits’ make perfectly adequate stand-ins for whatever multiplicity of identical entities is being counted/calculated/computed.
    Computers are nothing more than glorified abacuses – vastly accelerated, massively miniaturised, user-friendlied, now globally interconnected, electronically automated abacuses. ABACUSES. BEAN COUNTING MACHINES. A fact which George Gilder points out in ‘Life After Google’. Even IIT advocate, Christof Koch in ‘The Feeling of Life Itself’ says consciousness (that is real thinking) cannot be computed into being. As also among a growing number of others, does Sir Roger Penrose.
    As such – as beads on strings – ‘digits’ cannot be utilised as the means with which to ‘think’, the latter requiring real, true, bona fide ‘information’. Indeed, ‘digits’’ inadequacy in this ‘intelligent thinking’ regard obtains no matter how many of them one has at one’s disposal, nor how cleverly arranged they are, nor how large, powerful & globally interconnected are the machines & devices operating on them. Let alone ‘intelligently’.
    Starting with our own fingers & toes & moving on to scratch marks on prison walls, ‘digits’ (sic) are excellent instruments with which to count, calculate & compute, but, again, they are not units of information, & they cannot be utilised as aids to real, true bona fide ‘thinking’.
    ‘Thinking’ is ‘using information (not digits) to guide & direct action, such action as is executed in regard to whatever object/s &/or event/s initially gave rise to the information being so utilised’.
    On the other hand & by this definition of ‘thinking’ (as opposed to merely ‘counting’/‘calculating’/‘computing’) anything that is equipped to first take in some certain kind & amount of real information from its surrounds – temperature, humidity & barometric pressure, say – & then use this information to guide & direct those of its actions which it then executes strictly in regard to those very objects & events which initially gave rise to the information it is so using, is ‘thinking’ – not just counting.
    As several others have already noted, our own climate control systems think. As do our robovacs etc.
    Even though these systems genuinely ‘think’, they do not do so ‘intelligently’. Only ‘wholes’ can think intelligently. All of our own otherwise perfectly properly ‘thinking’ servomechanisms – all of our own humanly-invented & manufactured cybernetic machines, tools, systems, & devices – are what Richard Dawkins correctly labelled ‘extended phenotypes’. They are components of our own intelligent use of the information they routinely gather & use within their many & various operations, but they do not do so in their own regard.
    ‘Intelligence’ is ‘thinking’ in a fitness maximising/optimising manner. That is to say, ‘intelligence’ is using information to guide and direct action/behaviour in a fitness maximising/optimising manner, such action as is executed by the thinking/acting agent in regard to whatever object/s &/or event’s having given rise to the information being so used in the first instance’.
    Although it has been my most dubious fortune to have figured out ‘information’s’ correct ontological identity, I am not going to divulge it here in this YouTube comment (otherwise I’d have to kill you, as the saying goes), but I can most certainly assure you that once ‘information’s’ correct ontological identity is properly recognised, no great difficulty whatsoever attends the run-on tasks of further determining the ontological identities of all of the directly information-related phenomena such as ‘thought’, ‘mind’, ‘intelligence’ and ‘consciousness’ – & further, utilising information’s’ correct identity also enables us to clearly & easily distinguish between the two realms being conflated here – information & real thinking, and, digits & mere counting.

  40. @Torterra_ghahhyhiHd

    January 27, 2026 at 4:40 am

    actually, i love to build an intelligence without any propose and ambition. because every filosofy stuff etc we thin is right may is really not. could be so much of prejudice inflated by ego or fear of wasted effort, o fear to die, etc. we have so limited real deep understanding. that we assume die i not good a life is good. and if it is wrong the go to the opposite directions. our mind is still mindless compared to real deep meaning in the universe. we say we are aware, we are not aware of the real values even the moral value, etc is full of bugs. we just intent to approach. so every wrong conception on filosofical stuff that we put in the ia it will be exponentially, reflected in the derived way and calculus an integrals problem magnify in the world. keep it pragmatic, keep it simple. do not create firearms. do not realize any complex task like save the world, eliminate poverty or hunger, suppose to get a "mind", be "kind", "don't be kind" (this base is not solid and part of this construction is prejudiced itself in our filosofical conception), etc. "love", "hate".

Comments are closed.

Related
Minnesota Fraud Scandal EXPLAINED
Leeja Miller | April 18, 2026
Related
This anti-sex trafficking law could end internet freedom
Vox | April 18, 2026
Related
BOMBSHELL: Was The TRUMP Admin Leaking Intel To Russia???
christophertitustv | April 18, 2026

This area can contain widgets, menus, shortcodes and custom content. You can manage it from the Customizer, in the Second layer section.

 

 

 

Go to album
  • play_circle_filled

    92.9 : The Torch

  • play_circle_filled

    AGGRO
    'Til Deaf Do Us Part...

  • play_circle_filled

    SLACK!
    The Music That Made Gen-X

  • play_circle_filled

    KUDZU
    The Northwoods' Alt-Country & Americana

  • play_circle_filled

    BOOZHOO
    Indigenous Radio

  • play_circle_filled

    THE FLOW
    The Northwoods' Hip Hop and R&B

play_arrow skip_previous skip_next volume_down
playlist_play