OpenAI's Dan Roberts: Why AI Can Now Make Discoveries

Data Driven NYC
Data Driven NYCJun 4, 2026

Why It Matters

AI’s emerging capacity to generate and verify scientific insights promises to accelerate discovery cycles, turning massive compute into a collaborative research partner for academia and industry alike.

Key Takeaways

  • Reinforcement learning drives AI's ability to autonomously discover science.
  • OpenAI’s informal reasoning models solved Erdős conjecture without formal proof language.
  • Contrast: DeepMind uses formal Lean proofs; OpenAI relies on natural language reasoning.
  • Roberts’ physics background informs his view of AI as computational universe.
  • Scaling RL and test‑time compute accelerates AI’s role in mathematical breakthroughs.

Summary

In this interview, OpenAI researcher Dan Roberts explains how reinforcement learning and test‑time reasoning are enabling AI systems to tackle deep scientific problems, highlighted by recent breakthroughs on long‑standing Erdős conjectures. Roberts outlines the distinction between OpenAI’s informal, language‑model‑based approach—where models reason directly on natural‑language statements—and DeepMind’s formal‑proof strategy using the Lean theorem‑proving language. He emphasizes that both methods reflect a broader shift: AI is moving from merely executing tasks to autonomously exploring hypotheses, persisting through long, contrarian reasoning paths, and leveraging massive compute to generate novel insights. The discussion also touches on Roberts’ own journey from theoretical physics to AI, illustrating how concepts from quantum gravity and information theory shape his view of AI as a computational embodiment of physical laws. Ultimately, the conversation underscores a gradual, not abrupt, transition where AI increasingly augments scientific discovery, reshaping research workflows and accelerating progress across mathematics and physics.

Original Description

Are we witnessing the first real signs of AI becoming a scientist? In this episode of The MAD Podcast, Matt Turck sits down with Dan Roberts, lead of the Foundations of Reinforcement Learning team at OpenAI, to explore one of the biggest shifts happening in AI: the rise of reasoning models, test-time compute, and reinforcement learning as engines of scientific discovery. Dan brings a rare perspective - from theoretical physics, black holes, quantum information, and deep learning theory - to explain how models are learning to “think,” why language may be such a powerful foundation for intelligence, what recent AI math breakthroughs really mean, and whether we are beginning to see AI systems that can contribute to science itself.
Dan Roberts
OpenAI
Matt Turck (Managing Director)
FirstMark
Listen on:
00:00 Intro: AI's wild week in mathematics
01:21 What OpenAI's Foundations of RL team does
03:08 Dan's journey: from black holes and quantum gravity to frontier AI
07:04 Are AI systems becoming useful for real science?
08:21 The AI math moment: Erdős, OpenAI, DeepMind, and Anthropic
08:52 Why the OpenAI result was an act of exploration
10:25 OpenAI vs. DeepMind: informal reasoning vs. formal proof
12:13 RL 101: learning by doing, not just watching
15:10 Why reinforcement learning works
15:58 How RL breaks: sparse feedback and long-horizon tasks
17:03 RLHF: how human feedback shaped early language models
18:48 Move 37, self-play, and the search for novel strategies
22:16 Explore vs. exploit in scientific discovery
24:49 Why RL may now be "the cake," not the cherry on top
25:46 Why RL started working with large language models
27:29 Is RL "sucking supervision through a straw"?
28:47 Why language may be the grounding layer for intelligence
31:46 A contrarian take on the Bitter Lesson
32:41 What test-time compute actually is
34:50 How RL gives models the ability to think
35:40 Verifiable rewards, math, coding, and the messy real world
38:00 What physics can teach us about AI
42:08 Is there a thermodynamics of AI?
43:08 From Erdős problems to Einstein-level AI
45:16 Is AI already doing original science?
45:51 How far are we from AI automating AI research?
47:41 Why Dan is excited about the future of science

Comments

Want to join the conversation?

Loading comments...