Our new blog for science

We’re starting the OpenAI science blog to give scientists and academic researchers a place to share what they’re discovering with AI and how they’re doing the work. Our mission is to ensure that AGI benefits all of humanity, and accelerating scientific discovery is one of the most impactful ways we can work toward that goal.

We’ll publish work across a broad range of scientific disciplines, including:

• Research and discoveries: New scientific results and updates on ongoing research from OpenAI, our collaborators, and the broader scientific community.

• Methods and practical guidance: Examples of how scientists use our models for tasks such as scientific computing, data analysis, exploring literature, and planning experiments. We’ll share approaches others can adapt, including how to check the work and where a method falls short.

• Perspectives from the scientific community: Firsthand accounts from researchers using AI, what they are learning, and how these capabilities are affecting their disciplines. We’ll also discuss open questions about scientific practice and how to make these tools useful to more people.

Science expands our understanding of the world and makes advances possible, from better medicines to new sources of energy. Many of those advances begin with fundamental questions whose practical significance becomes clear only years later.

In recent months, we’ve seen an expanding range of scientific problems our models can address. In September, an internal OpenAI system produced a proof addressing the Navier–Stokes existence and smoothness problem⁠, a longstanding question about the equations describing fluid motion. In early October, we released a broad range of mathematical results⁠. Among them was a result on the quasi-Riemann hypothesis⁠(opens in a new window), with a result that sharpens estimates of how many prime numbers lie below a given number. Later that month, we shared results across mathematics, computer science, and physics⁠ concerning how waves concentrate, the limits of efficient computation, and the mathematical foundations of interlacing quantum field theory.

In biology, researchers are using AI to analyze complex datasets, connect findings across fields, and develop new hypotheses to test in the lab. Working with PerturbAI and Scripps Research⁠, thousands of agents powered by GPT‑Rosalind analyzed existing experimental data from millions of mouse brain cells in 32 minutes. By connecting the effects of gene disruptions with scientific literature, they identified possible biological mechanisms for scientists to test. Researchers are also developing their own approaches: César de la Fuente’s team uses ChatGPT and Codex⁠ to build algorithms that identify potential antibiotic molecules, which they then make and test in the lab.

These capabilities bring new possibilities and changes to scientific practice. In mathematics, systems that generate proofs raise questions about how new arguments are checked, explained, and incorporated into the field. In experimental sciences, exploring more hypotheses creates opportunities — and challenges — in deciding which experiments to pursue. Across disciplines, researchers are considering what these tools mean for training, attribution, and the skills their work requires. The implications will differ by field, and scientific communities should help shape how these tools are used.

We hope this blog helps more people understand how AI is changing science and how they can use it to explore new questions and make discoveries of their own.