StarZero All shows Try the podcast clipper →

Dan Roberts

Research Lead, OpenAI · 1 appearance on the record.

computed by AI from the episodes · how this works → · full disclaimer →

scientistauthorfounderacademic@danintheory ↗LinkedIn ↗danintheory.com ↗

He is the co-author of the textbook The Principles of Deep Learning Theory, which applies theoretical physics to neural networks and reinforcement learning. He previously earned a PhD in physics from MIT, conducted postdoctoral research at the Institute for Advanced Study, and co-founded Diffeo, which was acquired by Salesforce.

17statements → 7claims → 1claims resolved → 3.65/5average certainty → 2.29/5average debate potential → 4.1/5argument clarity · the sources →

1 supported 0 partly supported 0 contradicted 6 not checkable as stated how the 7 claims stand · each chip opens the sources

4 predictions · 3 assertions · 6 insights · 4 disclosures · every statement was checked. The predictions and assertions are the 7 claims: statements the public record can support or contradict. 1 is resolved, and 6 name no date, number or outcome precise enough to check. Everything else (opinions, insights, what ifs, disclosures) can never be settled by the record, so it carries no assessment.

The record, in short

What the tape says about how Dan argues and how the claims held up. Everything they said, and everything said about them, is in the tabs below.

Their most notable supported claim

Assertion Supported
ChatGPT disproved an Erdős conjecture using cross-disciplinary mathematical reasoning
“The big result was that this conjecture of this lower bound for the number of pairs that you can make is, is false. Not only is it false, it was false due to a really interesting connection to another field of mathematics.”
Dan Roberts Jun 4, 2026 ▶ 9:57 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries

Argument clarity: do they answer the question? how? →

4.1 / 5 directness 4.1 · coherence 4.3 · precision 3.8 · compression 3.4

redirected or did not address 1 of 12 assessed questions (8%). Watch them ▸

This is a score against a rubric. It is not a rank. Every host question → answer exchange is scored with names hidden on directness, coherence, precision and compression, 1–5 each, on meaning alone: disfluencies are ignored, and only raw unedited episodes count. This is the score that measures thought. Every scored exchange, scores shown → · The rubric and its checks →

How they sound: speaking style how? →

286 words/min while actually speaking · 14.3 um and uh per 1k words

Measured by listening to the audio itself: 8,060 words across 1 episode of raw-level tape, transcribed verbatim with every um and uh kept, each one attributed only where the alignment onto our timed stream is unambiguous. These are measurements of speaking style. We do not rank them: across this corpus, fluency and argument quality are nearly uncorrelated (ρ≈0.2), and smooth talking does not signal clear thinking. How it's measured →

Everything Dan Roberts said on the MAD Podcast that made the record, most notable first. Filter by type, assessment or year in the ledger →

Insight
Dan Roberts: AI scaling requires novel algorithms, not just pure compute
“It's not that scale is all you need. You need to also have good ideas to guide the scaling.”
Dan Roberts Jun 4, 2026 ▶ 31:56 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Insight
Dan Roberts: AI models do not experience discontinuous emergence or grokking
“You have these crazy, huge systems that have all sorts of interesting phenomena, and, you know, if you think about it the right way, they don't grok. There's just this nice continuity.”
Dan Roberts Jun 4, 2026 ▶ 41:47 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Insight
Pre-training models on language before reinforcement learning is the correct architecture
“Having the model have a prior of language and being able to like, think in language and then train on top of that, that seems like clearly the right. The right thing to do.”
Dan Roberts Jun 4, 2026 ▶ 31:16 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Assertion Not checkable as stated
Combining reinforcement learning with pre-training outperforms scaling pre-training alone
“If you were just trying to scale pre-training, you wouldn't get anywhere near as far as also trying to scale RL on top of pre-training, which is what we do now.”
Dan Roberts Jun 4, 2026 ▶ 32:05 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Assertion Supported
ChatGPT disproved an Erdős conjecture using cross-disciplinary mathematical reasoning
“The big result was that this conjecture of this lower bound for the number of pairs that you can make is, is false. Not only is it false, it was false due to a really interesting connection to another field of mathematics.”
Dan Roberts Jun 4, 2026 ▶ 9:57 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Disclosure
OpenAI plans to increasingly rely on reinforcement learning to scale intelligence
“When you have a lot of compute, you want to turn that compute into intelligence in a way that's useful, and RL is one way of doing it, and we just started doing it then, and we're going to do a lot more of it now.”
Dan Roberts Jun 4, 2026 ▶ 25:35 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Insight
Dan Roberts: AI scaling laws should be analyzed big-to-small
“The way to think about scaling and say, scaling laws is not small to big, but big to small.”
Dan Roberts Jun 4, 2026 ▶ 38:38 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Assertion Not checkable as stated
Current AI models lack research taste and problem formulation ability
“There's part of the scientific process. I think that the models haven't been imbued with yet. And I'm sure people are thinking about how to do that. You know, like what, trying to get to what is the right question as opposed to here's a well-defined thing and …”
Dan Roberts Jun 4, 2026 ▶ 44:53 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Prediction Not checkable as stated
Dan Roberts expects more AI-driven math and science breakthroughs within six months
“For the next six months. Like, I think we'll see more of these sorts of math and science breakthroughs.”
Dan Roberts Jun 4, 2026 ▶ 47:22 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Prediction Not checkable as stated
AI systems evolving into fully fledged scientists will be a gradual transition
“The, there's no sharp point, or I don't think there will be a sharp point where we'll say that systems didn't, weren't able to be useful for scientific, the scientific process to their fully fledged scientists. There'll be sort of a gradual shift.”
Dan Roberts Jun 4, 2026 ▶ 7:32 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Disclosure
OpenAI publishes most math research results using informal language settings
“Most of our results that we publicize, as far as I can think, are all in the informal setting.”
Dan Roberts Jun 4, 2026 ▶ 11:57 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Insight
Roberts: Powerful pre-trained models are necessary for effective RL and reasoning
“If you have a powerful enough pre-trained model, then it can start to do well at RL. It can start to like think at use test time compute to for instance, solve, solve math problems that it wouldn't otherwise be able to do.”
Dan Roberts Jun 4, 2026 ▶ 27:15 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Insight
Roberts: AI models improve performance by generating running thought tokens in language
“The natural way it thinks is in language. It's a language model, and so that's sort of this key insight that, that you can cause it to do better just by producing a thought process in, in, in token space, in, in language.”
Dan Roberts Jun 4, 2026 ▶ 34:01 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Prediction Not checkable as stated
OpenAI will release reinforcement learning products for consulting, banking, and legal
“I definitely think OpenAI will have amazing products that will be relevant in those domains, and some amount of RL will play a role in there.”
Dan Roberts Jun 4, 2026 ▶ 37:01 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Prediction Not checkable as stated
Continuous AI improvements render multi-year autonomous agent runs highly inefficient
“In, in general, we're not just going to like set up a system and let it think autonomously for eight years, if anything, because like the system's eight years. After will be so much more powerful that it probably doesn't make sense to let a system think for a …”
Dan Roberts Jun 4, 2026 ▶ 44:03 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Disclosure
OpenAI studied reinforcement learning reasoning models internally long before releasing o1
“So before we released a one and thinking reasoning models, we were studying this internally”
Dan Roberts Jun 4, 2026 ▶ 1:54 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries
Disclosure
OpenAI's RL team researches models two generations ahead of current releases
“And to do that, we need to make thinking models and some, somewhere along the way, we interact with that process. Usually at the earlier stage for models, you know, not the next model, but things that are like the next model or the next, next model.”
Dan Roberts Jun 4, 2026 ▶ 2:54 OpenAI's Dan Roberts: Why AI Can Now Make Discoveries

Appearances (1)

EpisodeDateSpeaking time
OpenAI's Dan Roberts: Why AI Can Now Make Discoveries Jun 4, 2026 37m
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 400 conversations transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more