OpenAI forms ‘OpenAI for Science’ team to speed physics, math discoveries
OpenAI just rolled out a new unit called OpenAI for Science. The idea? To speed up breakthroughs in physics and math by training AI that can actually reason about scientific problems.
Kevin Weil, who was just named vice-president of AI for Science, will be running the show. We don’t have a detailed roadmap yet, and it’s unclear how long it will take to see results, but the emphasis on computational discovery hints they might be building more specialized models. In my view, the team is trying to link modern machine-learning tricks with the careful methods scientists use every day.
This follows OpenAI’s recent push into applied research, yet they haven’t said much about budgets, hiring plans, or timelines. How will researchers? If the group manages to hand over tools that genuinely help physicists and mathematicians, we could see a noticeable bump in theoretical work.
Still, without clear metrics, we’ll have to wait and see how the program unfolds.
OpenAI is expanding its research efforts with a new program called OpenAI for Science, created to develop AI systems that can support scientific reasoning and speed up discoveries in physics and mathematics. The initiative is led by Kevin Weil, who has taken on the role of VP of AI for Science and posted about the program about a month ago. Now one of the first outside scientists to join the team is black hole researcher Alex Lupsasca, who will continue to hold his professorship at Vanderbilt University.
In an X thread, Lupsasca shared that his decision to join OpenAI was influenced by recent advances in the company’s technology. While he had previously considered AI to be far from capable of matching human researchers, his view changed after working with GPT‑5 Pro. In his paper "Why there is no Love in black holes", which describes new conformal symmetries in Kerr perturbations, Lupsasca found that GPT‑5 Pro was able to rediscover the central symmetry in about half an hour after minimal setup.
OpenAI’s “OpenAI for Science” program says it wants to speed up breakthroughs in physics and mathematics by building systems that reason like scientists. Kevin Weil is heading the effort, which was announced about a month ago. Already the team has brought on Alex Lupsasca, a black-hole expert from Vanderbilt, who will stay on the faculty while working with OpenAI.
The idea is simple enough: pair massive language models with tools that understand the specific math or physics problem, hoping to shave weeks off hypothesis generation and testing. What the announcement doesn’t show, though, is how those models will slot into a lab’s daily workflow, nor any numbers that prove they actually cut time. The whole thing will probably depend on whether the AI can cope with the exacting standards of formal proof or high-resolution simulation.
It’s also unclear if a partnership between external professors and OpenAI engineers will be enough to turn raw output into peer-review-ready results. As it stands, the program widens OpenAI’s reach, but we’ll have to wait and see if it really moves discovery forward.
Further Reading
More in Research & Benchmarks
Dell and NVIDIA host AI developer meetup in Bengaluru on deployment trade‑offs
Dell and NVIDIA have converged on Bengaluru for a hands‑on gathering aimed at the engineers and...
GPT-5.2 leads FrontierScience test, but falters on real research tasks
OpenAI’s latest benchmark, the FrontierScience test, pits its newest model, GPT‑5.2, against a...
OpenUSD and NVIDIA Halos Enhance Robotaxi Safety with Synthetic Data, SimReady
Robotaxi developers have long wrestled with the gap between virtual testing and real‑world...
Audio Dataset Valuable for Listening Models, Tackles Noise, Accents, Timing
Why does a single audio collection matter when the field is flooded with text‑heavy benchmarks?
Fastweb and Vodafone use LangGraph LLM Compiler to automate customer requests
Why does this matter? Because two of Italy’s biggest telecoms have moved from isolated chatbots to...
Common Questions Answered
What is the primary goal of the new 'OpenAI for Science' team?
The team's mandate is to build AI systems that can handle scientific reasoning and speed up discovery pipelines, specifically in physics and mathematics. This initiative aims to accelerate breakthroughs by combining large-scale models with domain-specific expertise.
Who is leading the 'OpenAI for Science' initiative at OpenAI?
Kevin Weil, the newly appointed vice-president of AI for Science, is leading the 'OpenAI for Science' effort. He announced the program about a month ago and is responsible for guiding the team's research direction.
Which external researcher has already joined the 'OpenAI for Science' team?
Black hole researcher Alex Lupsasca is one of the first outside scientists to join the 'OpenAI for Science' team. He will continue to hold his professorship at Vanderbilt University while contributing to the program's efforts.
What specific scientific domains is the 'OpenAI for Science' program targeting?
The program is explicitly focused on accelerating discoveries in physics and mathematics. Its goal is to develop AI systems capable of supporting scientific reasoning within these specific domains.