May 5, 2026 · 1h 31m · latent-space
🔬How GPT‑5 derived new results in theoretical physics and quantum gravity — Alex Lupsasca, OpenAI
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Theoretical physicist and OpenAI fellow Alex Lupsasca discusses how frontier AI reasoning models are revolutionizing theoretical physics by deriving groundbreaking new results in quantum field theory and quantum gravity. He details landmark discoveries in gluon and graviton scattering amplitudes while exploring the future of scientific taste, researcher workflows, and automated verification.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The hosts as informed peer 3.9 Guest teaching 4.3 Guest disagreement 0.3 The hosts pushing back 0.9
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:08:14 Dismissing the qualitative gap between human and machine creativity
When RJ suggests AI may merely be recombining existing knowledge rather than inventing new insights, Alex dismisses the distinction by questioning whether humans are anything more than recombination machines themselves.
Hardest push from the hosts ▶ 1:16:33 Challenging AI's ability to make unprompted conceptual leapsBrandon presses Alex on whether AI can truly discover physics without human priming, challenging him on whether a model cut off prior to 1904 could independently invent relativity or discover Kerr metric anomalies.
Biggest teaching moment ▶ 9:05 Masterclass on scattering amplitudes and force carriersAlex delivers an extensive, rigorous breakdown of how quantum field theory formulates probabilities from squaring complex amplitudes across particle colliders, completely commanding the conceptual framing.
The host holds their own ▶ 1:01:27 Brandon articulates the sociological problem of taste and fads in theoretical physicsBrandon demonstrates strong domain literacy by challenging how superhuman calculational power interacts with a high-energy physics landscape often dominated by aesthetic trends and non-empirical assumptions.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| From Skepticism to OpenAI: Lupsasca's AI Epiphany | 3 | 2 | 0 | 0 | Alex details his path from an AI skeptic to joining OpenAI after reasoning models replicated his research in minutes. The co-hosts validate his experience by drawing parallels to Andre Karpathy and Codex adoption in coding. | |
| Reconciling Relativity and Quantum Mechanics via Quantum Field Theory | 1 | 7 | 0 | 0 | Alex delivers an extensive pedagogical breakdown of how quantum field theory reconciles relativity with quantum mechanics via complex scattering amplitudes and helicity. The hosts remain in receptive listening mode, offering only brief clarifying prompts. | |
| Deriving Single-Minus Gluon Tree Amplitudes Using GPT Models | 5 | 6 | 0 | 1 | Alex explains how GPT discovered and proved concise formulas for single-minus gluon amplitudes that human theorists had struggled with for over a year. Host Brandon shows solid technical domain grasp by contextualizing tree versus loop expansions as polynomial approximations and noting dimensional analysis arguments. | |
| Generalizing Single-Minus Amplitudes to Gravitons and Quantum Gravity | 4 | 5 | 0 | 1 | Alex discusses generalizing the gluon amplitudes to gravitons in three weeks using directed matrix tree theorems with GPT Pro. RJ and Brandon frame questions around graviton spin definitions and Wigner representations. | |
| Impact of AI on Physics Education, Graduate Training, and Workflows | 4 | 3 | 0 | 1 | RJ questions how graduate physics training must evolve when students no longer need to execute tedious calculations themselves. Alex shares his perspective on how AI acts as an intellectual scout that drastically reduces confusion time. | |
| Scientific Taste, AI Reasoning Horizons, and Black Hole Symmetries | 5 | 4 | 1 | 2 | Brandon presses Alex on scientific 'taste' and whether AI might chase mathematical fads without empirical grounding. Alex counters by illustrating the edge of knowledge and recounting how GPT-5 independently deduced black hole tidal love number symmetries. | |
| Future Scientific Horizons, Automated Verification, and Paper Bottlenecks | 5 | 3 | 1 | 1 | Alex and the hosts explore the future bottlenecks of science, concluding that human verification and obsolete static paper formats are the main constraints. Alex notes that while formal verification like Lean was once deemed unnecessary for intuitive reasoning, AI paper flooding makes automated checking essential. |
Statements from this episode (13)
GPT-5 Reproduced Lupsasca's Best Physics Paper in 30 Minutes
“Then when GPT-V came out. It was able to reproduce one of my best papers that took me a very long time to come up with, in like, 30 minutes.”
Codex Wrote Complex SYK Physics Simulation in 10 Minutes
“Codex just wrote up a simulation of the SYK model. This is like a very technical thing in quantum mechanics and gravity. And like, yeah, a lot of research groups have been trying to run this simulation and it couldn't do it. And Codex did it in 10 minutes.”
AI Resolved Theoretical Physics Problem That Puzzled Experts for a Year
“AI has become superhuman, at least on certain tasks. And that's what led to these recent papers which maybe we should talk about that resolve a problem that was puzzling physicists for experts in the field for over a year, and they weren't able to resolve it a…”
ChatGPT Solved Open Physics Problem Before Collaborator's Flight Landed
“We decided to start working on it using AI a little bit before Andy was scheduled to come, like the week before. And in fact, using ChatGPT, we solved the problem before he even got off the plane.”
Internal OpenAI Model Proved Gluon Amplitude Formula in 12 Hours
“We had this Internal model that could think for a very long time and was extra strong in physics. So we gave it the whole problem from scratch without actually giving it this. We just formulated the problem in a very sharp way and asked the model to solve, to …”
ChatGPT Pro Derived All Math in Recent Quantum Gravity Paper
“It's a real solid result in quantum gravity that was done pretty much completely by an AI. With humans steering it and asking kind of the right questions, but all the math was derived by ChatGPT Pro, the public model you can access.”
Frontier AI Models Can Solve Six-Month Graduate Physics Starter Problems
“And I think the issue is that many such problems now, I would say these models can probably crush. Yeah. These are problems that we usually take again, you know, timescale for a theoretical physics paper is six months to a year. That's pretty typical.”
Physicists Can Use Parallel AI Chats to Scout Unknown Research Paths
“With AI, actually, you can launch 10 instances of chat and have each one try a different route and send it as a scout that moves very fast into the unknown, pushing outwards. And you can just very quickly get some feedback to see, okay, these approaches are no…”
Terry Tao Says AI Math Proofs Merely Cite Obscure References
“I talked to Terry Tao a couple of weeks ago at UCLA. We had an OpenAI event with IPAM, which is this Institute of Mathematics there. And I talked to Terry Tao and he said that in his view, all of the proofs that he's seen AI come up with in math, even the ones…”
ChatGPT Pro Generated Lupsasca's Exact Top Three Follow-Up Physics Questions
“You can take this page of this paper and you can feed it to ChatGPT Pro, say, like the best model we have out right now, and you can ask it, what should I do next? Give me the top three follow-up questions to ask based on this paper. I've done this experiment …”
Current AI Models Churn Out Papers Equal to Human-Written Ones
“I think we now have models that can really churn out papers that are as good as Hubert written papers.”
Interactive LLMs Will Replace Static Scientific Papers Within 20 Years
“If you ask me, would I be confident that in 20 years we'll have these sort of like static documents in which we publish our results as papers? I would think not. Like that doesn't seem like the best thing we could be doing. Maybe some kind of interactive paper…”
AI Proof Generation Shifts Human Scientific Burden to Automated Verification
“Now that we're in this regime where you can just get ChatGPT to tackle thousands of questions at the same time, and it will return proofs for a significant fraction of them. Now, actually the onus is back on the humans to verify all the outputs. And so, yeah, …”