New Research Shows How AI Could Transform Math, Physics, Cancer Research, and More

New study shows how AI could change math, physics, cancer research and more

TO Denis Ellis Bechard edited by Clara Moskowitz

A new report from OpenAI and a team of third-party scientists shows how GPT-5, the company's latest large artificial intelligence language (LLM) model, can help with research from black holes to cancer-fighting cells and math puzzles.

Each chapter of the article presents case studies: a mathematician or physicist in a difficult situation, a doctor trying to confirm the results of laboratory tests. They all ask GPT-5 for help. Sometimes the LLM gets it wrong. Sometimes he finds a faster path to an already known result. But in other cases, with careful human guidance, it helps expand the boundaries of what was previously known.

In one experiment looking at the behavior of waves around black holes, GPT-5 performed mathematical calculations to independently produce results that had previously been proven correct, showing that it was capable of performing this level of science. In another nuclear fusion project, GPT-5 developed a model that accelerated research. “The ability of artificial intelligence to significantly reduce the time required for coding—compressing what would traditionally take days for an author into minutes—has enormous implications for research practice,” says Floor Bruckgaarden, an astronomer at the University of California, San Diego, who was not involved in the study.

About supporting science journalism

If you enjoyed this article, please consider supporting our award-winning journalism. subscription. By purchasing a subscription, you help ensure a future of influential stories about the discoveries and ideas shaping our world today.

In another case, researchers studying immune cells used GPT-5 to interpret their data, and his explanation was consistent with results the lab had already confirmed. “GPT-5 Pro can act as a true mechanistic co-investigator in biomedical research, condensing months of reasoning into minutes, uncovering non-obvious hypotheses, and directly shaping experimentally testable strategies,” Derya Unutmaz, a doctoral researcher leading the project, wrote in the paper.

The paper also reports several new mathematical discoveries supported by GPT-5. Under the guidance of human experts, he solved a long-standing problem posed in 1992 by mathematician Paul Erdoeds. It also led to a clearer rule showing the limitations of how computer systems make decisions; discovered another rule for how certain small patterns appear within branching diagrams; and found a way to discover secret structures in the network as it grew. The discoveries are modest but appear genuine, and each has been verified by human mathematicians.

“I've never seen anything so impressive [in math] “I suspect the MSc will revolutionize the way we create, test and improve theories,” says Ryan Foley, an astrophysicist at the University of California, Santa Cruz, who was not involved in the study. However, he cautions that AI tools still require a lot of guidance: “People are creative; The AI ​​is responsive. However, the pace of discovery should accelerate quickly.”

Prithviraj Ammanabrolu, a computer scientist at the University of California, San Diego, who was not involved in the study, notes that the published work is more of a series of case studies than a scientific article because it does not provide enough detail to replicate the experiments and does not offer counterfactual analyzes that include different approaches. Despite these limitations, AI's ability to assist in research “is still way ahead of what was possible even a year ago, so the pace of progress is quite rapid,” he says. “This shows the future potential to allow scientists to precisely blend relevant previous results and gain new insights in new ways.”

One of GPT-5's strengths is its ability to search huge volumes of scientific literature. For a mathematical problem listed as unsolved on the Internet, the solution was found in a paper from the 1980s. In another case, several lines were found in a German newspaper from the 1960s that solved the problem. He easily overcame language barriers and differences in writing style between mid-century and modern mathematical writing.

This may all sound like the scientific genius of GPT-5, but the authors of the paper make it clear that this is not the case. Rather, in skillful hands, it is a fast and tireless assistant, who has read an impossible number of articles and is not afraid to redo the calculation. But human judgment is not necessary, they emphasize. The researchers also found that it is clearly flawed and can distort references, creating hallucinations of non-existent articles or not crediting real authors.

“The human experience is still critical,” Broekgarden says. But AI “can perform a variety of tasks — collating data, summarizing scientific articles, and even performing complex calculations — that previously required researchers a lot of time and effort.”

The many ways in which AI will impact research remain to be seen. New artificial intelligence models are released every few months. If the general-purpose chatbots that challenged high school math two years ago can now detect hidden structures in black hole waves and suggest new approaches to cell therapy, who knows what their successors will achieve?

It's time to stand up for science

If you liked this article, I would like to ask for your support. Scientific American has been a champion of science and industry for 180 years, and now may be the most critical moment in that two-century history.

I was Scientific American I have been a subscriber since I was 12, and it has helped shape my view of the world. science always educates and delights me, instills a sense of awe in front of our vast and beautiful universe. I hope it does the same for you.

If you subscribe to Scientific Americanyou help ensure our coverage focuses on meaningful research and discovery; that we have the resources to report on decisions that threaten laboratories across the US; and that we support both aspiring and working scientists at a time when the value of science itself too often goes unrecognized.

In return you receive important news, fascinating podcastsbrilliant infographics, newsletters you can't missvideos worth watching challenging gamesand the world's best scientific articles and reporting. You can even give someone a subscription.

There has never been a more important time for us to stand up and show why science matters. I hope you will support us in this mission.