What Nathan Lambert thinks about AI alignment
Machine-learning researcher on open language models; writes the Interconnects newsletter and the RLHF Book, after leading post-training at Ai2.
Nathan Lambert did not write this page.
We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.
Filter AI alignment · Newest first
29 more topics
Only Nathan Lambert on AI alignment. See everyone on AI alignment
10 dated positions, 2025 to 2026, in their own words. Our reading of what Nathan Lambert has said — not written or endorsed by them.
-
Their wordsAnd the frontier models being a bit ahead in these companies that have time to do research on risks on closed models before the open weight models come in 5 to 8 months is a good balance.
↗Open Language Models: Overview, Open Source History, Distillation, Risks, & Chinese Labsyoutube.com 16th of 19 in this recording
-
Their wordsAnd I expect this to be a 2-to-5-year transition where there are new risks and new things that society has to overcome.
↗Open Language Models: Overview, Open Source History, Distillation, Risks, & Chinese Labsyoutube.com 17th of 19 in this recording
-
Their wordsThese are one of the biggest misreads on the Chinese labs is that the Chinese labs have a risk posture. And they don't want to cause harm. But they're operating in a different ecosystem
↗Open Language Models: Overview, Open Source History, Distillation, Risks, & Chinese Labsyoutube.com 18th of 19 in this recording
- 3 weeks earlier
-
Their wordsThe leading proponents of this new wave of open-weight fear mongering have effectively been making a falsifiable prediction — that the current open-weight models are going to cause substantial, not seen before AI harms by crippling our cyber infrastructure. Based on all the evidence I have, which I’ve unpacked above, they’re set up to be wrong.
- 2 weeks earlier
-
Their wordslossy self-improvement remains my baseline on the trajectory of progress, and the increased discussion of extinction risk seems very misplaced
↗Why I still haven’t bought into true RSIinterconnects.ai 3rd of 3 in this piece
- 5 weeks earlier
-
Their wordsThe capability diffusion is determined by the lowest common denominator.
↗GLM-5.3: How Chinese labs keep stride with the frontierinterconnects.ai 2nd of 3 in this piece
-
Their wordsAt the end of the day, this type of safety barely matters when true open-weights are coming.
↗GLM-5.3: How Chinese labs keep stride with the frontierinterconnects.ai 3rd of 3 in this piece
- 18 months earlier
-
Their wordsAnd the important thing to say is that no matter how you want the model to behave, these RLHF and preference-tuning techniques also improve performance. So, on things like math evals and code evals, there is something innate to these, what is called contrastive loss functions.
↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 22nd of 44 in this recording
-
Their wordsAnd these reasoning behaviors emerge naturally. So these things like, "Wait, let me see. Wait, let me check this. Oh, that might be a mistake." And they emerge from only having questions and answers.
↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 23rd of 44 in this recording
-
Their wordsAnd humans are actually very good at reading or judging between two things versus... This goes back to the core of what RLHF and preference tuning is that it's hard to generate a good answer for a lot of problems, but it's easy to see which one is better.
↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 41st of 44 in this recording