How AI Transformed the World’s Largest Physicists Conference

I’m seated in the auditorium at the American Physical Society Global Physics Summit, the largest annual gathering of physicists worldwide, with 14,000 researchers attending this year in Denver, Colorado. We gather to hear prominent scientists present their groundbreaking research, yet many are now turning to artificial intelligence (AI) for clarification on complex topics.

As the presentations progressed, I frequently noticed an AI chatbot displayed on my laptop screen. These AI chatbots are utilized to simplify complex concepts. Questions like, “What are the benefits of transmon qubits?” and “Can you explain spintronics?” are addressed with rapid, bulleted emoji responses.

AI chatbots have shown promise in educational settings, but whether they can contribute significantly to real-world physics research remains a hot topic at conferences, sparking discussions in talks, intersessions, and networking events.

In a recent presentation, Harvard University researcher Matthew Schwartz highlighted that Anthropic’s Claude chatbot can tackle advanced physics problems with proficiency comparable to early-stage PhD students. Schwartz, who co-authored a study in January focused on quantum field theory, shared that collaborating with Claude sped up research that would typically take two years with human students.

He argues that AI could fundamentally change theoretical physics, stating he will no longer work with students who resist using AI tools. Schwartz believes that AI advancements could solve longstanding challenges in physics, such as harmonizing quantum mechanics with Einstein’s theory of general relativity, within the next five years. He metaphorically expressed that working with Claude made him feel akin to Einstein, proposing the idea of “10,000 Einsteins.”

However, Schwartz’s perspective represents an extreme viewpoint. CUNY professor Savannah Thais maintained that it’s premature to gauge how much technology will transform physics, emphasizing the AI’s capacity to generate plausible-sounding science without guaranteeing accuracy. Critical assumptions in particle physics can often obscure the validity of results.

During a session, Rachel Burley from the American Physical Society noted the initial enthusiasm over AI tools assisting physicists with writing and publishing scientific papers had quickly led to an overwhelming increase in journal submissions, straining the peer review process.

A recurring question from both formal presentations and informal dialogues was: As AI evolves, what roles will remain for human researchers? Matthew Ginsburg, a former physicist with extensive experience in AI at Google DeepMind, suggested that while AI may offer consensus expert opinions, innovation arises from researchers willing to challenge conventional understanding and pose unexpected questions.

Schwartz speculated that human physicists will primarily focus on setting research priorities based on interest and significance. He expressed concern that the transition could lead to complications before improvements manifest, stating, “It’s remarkable, yet slightly concerning.”