Fully autonomous robots are much closer than you think – Sergey Levine
Related videos
Chip design from the bottom up – Reiner Pope
Dwarkesh Patel
25.6k views
How GPT-5, Claude, and Gemini are actually trained and served – Reiner Pope
Dwarkesh Patel
55.4k views
Noem Sex Worker Leak Is Worse Than You Think
Philip DeFranco
114.5k views
What are we scaling?
Dwarkesh Patel
22.3k views
Sarah Paine – Why Russia Lost the Cold War
Dwarkesh Patel
136.4k views
Ilya Sutskever – We're moving from the age of scaling to the age of research
Dwarkesh Patel
55.6k views
Sarah Paine — How Russia sabotaged China's rise
Dwarkesh Patel
102.1k views
Some thoughts on the Sutton interview
Dwarkesh Patel
77.0k views
Richard Sutton – Father of RL thinks LLMs are a dead end
Dwarkesh Patel
435.9k views
It's So Much Worse For The UK Than You Think
StevenCrowder
250.3k views
Robotics Foundation Models Forecast Autonomous Household Operation by 2030
Achieve clarity on the technological roadmap accelerating the robotics revolution. Sergey Levine shares his 2030 median forecast for fully autonomous household robots.
Short Summary
- A median forecast of 2030 for widespread deployment of robots capable of running households autonomously.
- Robotics benefits from advancements in VLA models, leveraging prior knowledge from LLMs faster than initially expected.
- The focus shifts from theoretical research to industrial-scale deployment resembling "The Apollo Program."
- Embodied AI leverages physical interaction, providing stronger grounding signals than text alone, setting the stage for a self-improving data flywheel.
This discussion explores the near-term path to a general-purpose robotics explosion, detailing the necessary synthesis of known AI techniques (VLA models, imitation, RL) and the industrial effort required for scaling data collection and achieving physical robustness. We analyze why robotics might scale faster than self-driving cars and discuss the crucial role of physical interaction in grounding abstract knowledge.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
Chip design from the bottom up – Reiner Pope
Dwarkesh Patel
25.6k views
How GPT-5, Claude, and Gemini are actually trained and served – Reiner Pope
Dwarkesh Patel
55.4k views
Noem Sex Worker Leak Is Worse Than You Think
Philip DeFranco
114.5k views
What are we scaling?
Dwarkesh Patel
22.3k views
Sarah Paine – Why Russia Lost the Cold War
Dwarkesh Patel
136.4k views
Ilya Sutskever – We're moving from the age of scaling to the age of research
Dwarkesh Patel
55.6k views
Sarah Paine — How Russia sabotaged China's rise
Dwarkesh Patel
102.1k views
Some thoughts on the Sutton interview
Dwarkesh Patel
77.0k views
Richard Sutton – Father of RL thinks LLMs are a dead end
Dwarkesh Patel
435.9k views
It's So Much Worse For The UK Than You Think
StevenCrowder
250.3k views
Top Comments (10)
Interesting impedance mismatch between the lay generalist and the specialized technologist. The former wants predictions, timelines and macroeconomic impact assessments, the latter is thinking, "can I solve the next incremental goal?"
Balanced AI development is so important. Reminds me of Pneumatic Workflow's human-in-the-loop approach. Always good to keep humans involved!
Sergey’s RL lectures on YT are the best I’ve found.
Sergey obviously left his crystal ball at home.
Sergey could be right, but I don't think a founder of a robots company is ever gonna say it will take 30 year at least. He always will be positive because that's in his company best interest.
Dwarkesh: Asks any question Sergey: That's a very hard question to answer
Robotics are one of our best bets for productivity improvement. I work in manufacturing and it's great to see their improvements each year
The whole podcast Sergey was being badgerred to predict when would he solve robotics, while they just started their own research.
He was my Professor of DEEP RL course at Berkeley 7 years ago, when LLMs were just emerging onto the scene. He is a prodigy.
dwarkesh: what direction is the restaurant? Sergey: well, it’s hard to say. there’s north, south, east, and west.
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
Interesting impedance mismatch between the lay generalist and the specialized technologist. The former wants predictions, timelines and macroeconomic impact assessments, the latter is thinking, "can I solve the next incremental goal?"
Balanced AI development is so important. Reminds me of Pneumatic Workflow's human-in-the-loop approach. Always good to keep humans involved!
Sergey’s RL lectures on YT are the best I’ve found.
Sergey obviously left his crystal ball at home.
Sergey could be right, but I don't think a founder of a robots company is ever gonna say it will take 30 year at least. He always will be positive because that's in his company best interest.
Dwarkesh: Asks any question Sergey: That's a very hard question to answer
Robotics are one of our best bets for productivity improvement. I work in manufacturing and it's great to see their improvements each year
The whole podcast Sergey was being badgerred to predict when would he solve robotics, while they just started their own research.
He was my Professor of DEEP RL course at Berkeley 7 years ago, when LLMs were just emerging onto the scene. He is a prodigy.
dwarkesh: what direction is the restaurant? Sergey: well, it’s hard to say. there’s north, south, east, and west.