How artificial intelligence shook up the world’s largest meeting of physicists

I’m sitting in a lecture hall and there’s a sight in front of me that I’m still getting used to. I’m at American Physical Society Global Physics Summitthe largest annual meeting of physicists in the world, attended by 14,000 researchers this year in Denver, Colorado. We’ve all come to hear the world’s leading scientists talk about their work—yet many people are turning to artificial intelligence to help explain what we’re actually hearing.

As the conversations continue, I keep catching flashes of laptop screens showing AI chatbots being asked to put concepts into more understandable terms. “What are the advantages of transmon qubits?” “Explain spintronics to me. “What is a two-tier system?” AIs provide information instantly and use emoticons as bullet points.

While AI chatbots have proven their usefulness in lecture halls, whether they can help conduct real physics research is one of the hottest topics of the conference, discussed in every forum, from the talks themselves to the breakout sessions and alumni reception.

In one presentation, Matej Schwartz at Harvard University reported that Anthropic’s Claude chatbot can solve advanced physics problems as efficiently as a student in the early stages of a doctoral degree program. In January, Schwartz is a co-author of the study in quantum field theory by working with Claude for about two weeks. Surprisingly, he estimated that it would take about two years to complete the same research in collaboration with a student.

He believes that AI puts theoretical physics “on the chopping block”. Schwartz said that he no longer mentors students who do not want to work with AI tools, and believes that all the problems currently plaguing fundamental physics, such as combining quantum theory with Albert Einstein’s theory of general relativity, will be solved by AI in about five years. Working with Claude made him feel like Einstein himself—and like everyone could become Einstein’s equivalent, he said. His talk was called “10,000 Einsteins”.

Schwartz represents the extreme end of the spectrum. Savannah Thais at the City University of New York, said in her presentation that it’s too early to tell how transformative the technology will be for physics. She pointed out the fact that AI is good at producing plausible-sounding science, but there is no foolproof method for determining whether it is correct. Many steps are usually hidden from researchers, and the underlying assumptions made in, for example, particle physics can lead to less accurate results.

Rachel Burley at the American Physical Society said in a presentation that there was early optimism about how AI tools could help physicists write and publish scientific papers, but the subsequent explosion of journal submissions put pressure on the peer review system.

The question that loomed over these presentations and more informal conversations is what will be left for humans as AI advances. Matthew Ginsburga former physicist with decades of experience working on AI, including Google’s DeepMind, said AI provides consensus expert opinion, while scientific breakthroughs can come from researchers willing to go against the grain or ask unexpected questions.

Schwartz assumed that human physicists would be left with the task of creating taste, of determining which problems were the most interesting and meaningful. “My concern is that some things may get worse before they get better,” Schwartz said. “It’s amazing and also a little scary.

topics: