Nathan Lambert
speaker
1,814 appearances
3 recordings
2 series
first heard Feb 2025
last heard 1 Feb
Nathan Lambert’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 2 in all, peaking in Feb 2026 with 1.
Appearances
So you can essentially train your RL algorithm where compaction is an action, where it shortens the history, and then the problem formulation will be, I want to keep...
the maximum evaluation scores that i have gotten while the model compacts its history to the minimum length because then you have the minimum amount of tokens that you need to do this kind of compounding auto-aggressive prediction so there's actually pretty nice problem setups in this where the like these agentic models learn to use their context in a different way than just plow forward
Yeah.
All this scaling thing.
Like the reason we get the quad 4.5 Sonnet model first is because you can train it faster and you're not hitting these compute walls as soon.
And they can just try a lot more things and get the model faster, even though the bigger model is actually better.
I'm excited about the robotic learning space because I think it's collectively getting supercharged by all the excitement and investment in language models generally, where they're getting the infrastructure for training transformers, which is a general modeling thing, is becoming...
world-class industrial tooling where anything wherever that was a limitation for robotics it's just like way better there's where more compute and then on top of like they take these language models and use them as kind of central units where you can do interesting explorative work around something that kind of already works and then i see it emerging as like
kind of like we talked about Hugging Face Transformers and Hugging Face.
I think when I was at Hugging Face, I was trying to get this to happen, but it was too early.
It's like these open robotic models on Hugging Face and having people be able to contribute data and fine-tune them.
I think we're much closer now that the investment in robotics and I think self-driving cars is related and enables this, where it's like once you get to the point where you can have this sort of ecosystem where somebody can download a robotics model and maybe fine-tune it to their robot or
share data sets across the world and there's some data there's some work in this area like rtx i think is a few years ago where people are trying to do that but i think once they have this ecosystem it'll look very different and then this whole post chat chat gpt boom is putting more resources into that which i think is a very good area for doing research
problems you never thought you'd have to solve when you're just thinking about the general robot learning problem i'm so bearish on in-home learned robots for consumer purchase i'm very bullish on self-driving cars and i'm very bullish for robotic automation eg like amazon distribution where amazon has built whole new distribution centers designed for robots first rather than humans
There's a lot of excitement in AI circles about AI enabling automation and mass-scale manufacturing.
And I do think that the path to robots doing that is more reasonable, where it's like a thing that is designed and optimized to do a repetitive task that a human could conceivably do but doesn't want to.
But it's also going to take a lot longer than people probably predict.
I think the leap from...
ai singularity to we can now scale up mass manufacturing in the us because we have a massive ai advantage is one that is troubled by a lot of political and other challenging problems let's talk about timelines uh specifically timelines to agi or asi is it fair like as a starting point to say that nobody really agrees on the definitions of agi and asi
I kind of think there's a lot of disagreement, but I've been getting pushback where a lot of people kind of say the same thing, which is like a thing that can reproduce most digital economic work.
Showing 801–820 of 1,814 · page 41 of 91
← Previous
Next →