🔬Doing Vibe Physics — Alex Lupsasca, OpenAI
Get the full intelligence
Search transcripts, export clips, track mentions, and explore all topics from “🔬Doing Vibe Physics — Alex Lupsasca, OpenAI” inside PodZeus.
In this episode of Latent Space: The AI Engineer Podcast, host Brandon welcomes Alex Lupsasca, a professor at Vanderbilt University and fellow at OpenAI, to discuss the transformative role of AI in theoretical physics. Lupsasca recounts his journey from skepticism about AI’s capabilities in complex physics calculations to becoming a full-time researcher at OpenAI, where he has leveraged advanced language models like GPT-5.2 Pro to solve long-standing problems in quantum field theory, including proving that single-minus gluon tree amplitudes are non-zero—a result that overturned decades of textbook assumptions. The breakthrough was achieved through a collaborative 'vibe physics' approach, where AI acted as a creative partner, generating insights and simplifying complex calculations that had stumped experts for over a year. This success was replicated in a follow-up paper on graviton amplitudes, demonstrating AI’s ability to generalize across different physical systems. Lupsasca emphasizes that AI is not just a tool for computation but a collaborator that accelerates discovery, reduces confusion, and redefines how physicists approach research. He also raises critical questions about the future of scientific training, peer review, and the nature of creativity in science, urging the academic community to adapt to a new era where AI superpowers human intellect. The episode concludes with a forward-looking vision: AI is no longer a novelty but a central force in scientific progress. Lupsasca argues that the real bottleneck now is not computation, but verification and the ability to ask the right questions—skills that remain uniquely human. He envisions a future where papers evolve into interactive, AI-powered documents, and where the next generation of scientists must learn to steer AI with 'taste' and judgment. The takeaway is clear: we are at a pivotal moment in science, where AI is not just assisting researchers but actively shaping the frontier of knowledge. The message is urgent: pay attention, adapt, and embrace the new paradigm before it reshapes everything.
AI has become superhuman in specific scientific tasks, solving problems in theoretical physics that stumped experts for over a year.
The 'vibe physics' approach—using AI as a creative collaborator—can accelerate discovery by generating insights, simplifying complex calculations, and reducing researcher confusion.
AI is not just a calculator but a partner in scientific reasoning, capable of proposing conjectures and even deriving proofs, though human oversight remains essential.
The future of scientific research will likely shift from static papers to interactive, AI-enhanced documents that allow dynamic exploration of results.
The biggest challenge now is not computation, but verification and developing the 'taste' to ask the right scientific questions—skills that remain human-centric.
The AI Threshold: From Skepticism to Superhuman Physics
“I thought, oh my God, this changes everything. It's the most important discovery in my lifetime.”
Solving the Gluon Amplitude Puzzle: A Year-Long Mystery Cracked in Days
“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 ChadGPT, we solved the problem before he even got off the plane.”
From Chaos to Simplicity: AI's Role in Uncovering the Park-Taylor Formula for Gluons
“Instead of having this factorial growth which is super exponential... here it's actually linear. If you double the number of particles, you only double the number of terms.”
Extending to Gravity: The Graviton Paper and the Power of Generalization
Using the same AI framework, the team extended the result to gravity, solving the graviton amplitude problem in just days—showing AI’s ability to generalize across physical systems with different mathematical structures.
The New Reality: AI as a Scientific Collaborator and the Future of Research
Lupsasca reflects on how AI has changed his workflow—reducing confusion, enabling rapid exploration of multiple research paths, and shifting the role of the physicist from calculator to question-asker and verifier.
“I thought, oh my God, this changes everything. It's the most important discovery in my lifetime.”
“If you just extrapolate that into the future, imagine where we're going to be in six months or a year. I think it's kind of surreal to live through this time, but it's really happening.”
“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 ChadGPT, we solved the problem before he even got off the plane.”
Hosts
Guest
OpenAI
organization
Alex Lupsasca
person
GPT-5
product
ChadGPT
product
GPT-5.2 Pro
product
Alfredo Guevara
person
David Skinner
person
Andrew Strominger
person
Kevin Weill
person
Vanderbilt University
organization
Moonlake: Causal World Models should be Multimodal, Interactive, and Efficient — with Chris Manning and Fan-yun Sun
Latent Space: The AI Engineer Podcast • 1h 6m • 4/2/2026
Marc Andreessen introspects on The Death of the Browser, Pi + OpenClaw, and Why "This Time Is Different"
Latent Space: The AI Engineer Podcast • 1h 16m • 4/3/2026
Extreme Harness Engineering for Token Billionaires: 1M LOC, 1B toks/day, 0% human code, 0% human review — Ryan Lopopolo, OpenAI Frontier & Symphony
Latent Space: The AI Engineer Podcast • 1h 12m • 4/7/2026
Notion’s Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future — Simon Last & Sarah Sachs of Notion
Latent Space: The AI Engineer Podcast • 1h 17m • 4/15/2026
🔬 Training Transformers to solve 95% failure rate of Cancer Trials — Ron Alfa & Daniel Bear, Noetik
Latent Space: The AI Engineer Podcast • 1h 25m • 4/20/2026
Get the full intelligence
Search transcripts, export clips, track mentions, and explore all topics from “🔬Doing Vibe Physics — Alex Lupsasca, OpenAI” inside PodZeus.
Start discovering podcast insights today
Start with a 7-day trial and explore a growing catalog of popular podcasts. No credit card required.
No credit card required • 7-day trial • Cancel anytime
