What Jeff Dean thinks about LLMs

@jeff-dean · 22 positions · On ppll ↗

Chief Scientist at Google DeepMind and Google Research, at Google since 1999, where he co-created MapReduce, Bigtable, TensorFlow and the TPU.

Jeff Dean 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 LLMs · Newest first

Topic

All 22AI agents 4coding agents 3energy 3LLMs 3scaling laws 3semiconductors 2taste 2AI and jobs 1AI and science 1

4 more topics

design 1protein structure prediction 1recursive self-improvement 1startups 1

Sort

NewestOldest

Only Jeff Dean on LLMs. See everyone on LLMs

3 dated positions, 2026, in their own words. Our reading of what Jeff Dean has said — not written or endorsed by them.

3 positions so far — this page is not yet offered to search engines.

1. 30 Jul 2026

   Our readingWhat sits in a model's context is far clearer to it than its training data, which is trillions of tokens stirred into a soup of parameters.

   Their wordsAnd the nice thing about that is that information is really clear to the model, unlike the training data the model was trained on where it's all kind of like trillions of tokens stirred together into a soup of of hundreds of billions or trillions of parameters, but it's all less clear than the actual context uh that the model sees directly for this particular problem or uses use case.

   ↗Jeff Dean: The 1% Rule for Building in AIyoutube.com 8th of 22 in this recording

3. 30 Jul 2026

   Our readingA startup's durable advantage over a general model is having visibility into data the general model will never see.

   startups

   Their wordsYeah, I mean I think uh sometimes it's uh a product that you build that might have access to particular kind of data that the underlying model might not the a general model. So it might be you're building something to help users organize all their own personal information and the model won't necessarily have access to that. And so there you can have a big advantage because all of a sudden your model has visibility or your product has visibility into important data.

   ↗Jeff Dean: The 1% Rule for Building in AIyoutube.com 14th of 22 in this recording

5. 30 Jul 2026

   Our readingToday's models see a thousand times more data than an eighteen-year-old and are no better for it, so far more data-efficient algorithms must be possible.

   Their wordsIf you think about our large scale models today, they probably see a thousand times as much data as a human does by the age of 18. Yet, the human by the age of 18 is better in a lot of things and, you know, on par uh with those frontier models that have seen way more data. So could you come up with much more data efficient systems that can learn continuously learn from their own actions?

   ↗Jeff Dean: The 1% Rule for Building in AIyoutube.com 22nd of 22 in this recording