People who’ve both spent time in academia and AI seem to believe that things could change faster than outsiders expect.
Jared Kaplan, co-founder of Anthropic and former theoretical physicist, has made a striking prediction: there’s a 50% chance that theoretical physicists will mostly be replaced by AI within two to three years. His forecast suggests that AI systems will soon generate papers rivaling the work of luminaries like Nima Arkani-Hamed and Ed Witten—essentially autonomously.
“I would give like a 50% chance that in two or three years, theoretical physicists will mostly be replaced with AI,” Kaplan said in an interview. “Brilliant people like Nima Arkani-Hamed or Ed Witten, AI will be generating papers that are as good as their papers pretty autonomously.”
From Physics to AI: A Brain Drain Story
Kaplan’s credentials lend weight to his assessment. As a doctoral student at Harvard University in the 2000s, he worked alongside renowned theorist Nima Arkani-Hamed, helping to open new directions in amplitude research that physicists actively pursue today. After completing his PhD in Physics from Harvard (2005-2009) and a postdoctoral position at SLAC National Accelerator Laboratory (2009-2013), Kaplan eventually became a professor at Johns Hopkins University in 2012.
But in 2019, he left academia entirely. His reason? AI’s trajectory seemed too consequential to ignore.
“I started working on AI because it seemed plausible to me that AI was going to make progress faster than almost any field in science historically,” Kaplan explained. He viewed AI as potentially “the most important thing to happen while we’re alive, maybe one of the most important things to happen in the history of science. And so it seemed obvious that I should work on it.”
This brain drain appears to be real and accelerating. Kaplan joined OpenAI as a research consultant in 2019 before co-founding Anthropic, the company behind the Claude chatbot, where he continues to shape the frontier of AI development.
The Implications for Long-Term Planning
Kaplan’s prediction carries stark implications for how we think about scientific infrastructure and planning. When asked about the future of particle physics, he suggested that traditional long-term planning may be obsolete.
“I think that it’s kind of irrelevant what we plan on a 10-year timescale, because if we’re building a collider in 10 years, AI will be building the collider; humans won’t be building it,” he said. “So planning beyond this couple-year timescale isn’t really something I think about very much.”
A Broader Pattern Among AI Leaders
Kaplan’s timeline aligns with increasingly aggressive predictions from other AI leaders who’ve witnessed the technology’s rapid advancement firsthand.
Anthropic CEO Dario Amodei recently suggested that AI could compress 100 years of scientific progress into just 5-10 years. Google DeepMind’s David Silver believes AI mathematicians could fundamentally transform mathematics as a field. DeepMind co-founder Demis Hassabis has argued we could solve all diseases within 10-15 years through AI. Even former Google CEO Eric Schmidt recently predicted that AI could create superhuman mathematicians and coders within 1-2 years.
What distinguishes these predictions from typical tech hype is the credentials of those making them. These aren’t peripheral observers—they’re people building the technology, many with deep scientific backgrounds who understand both the domains being disrupted and the capabilities of current AI systems.
Questions That Remain
Kaplan’s 50% probability leaves substantial uncertainty. Will theoretical physics truly be automated within 2-3 years, or will the field prove more resistant to AI disruption than anticipated? Can AI systems truly match the creative intuition of elite physicists, or will they serve as powerful tools that augment rather than replace human researchers?
The timeline is aggressive, even by the standards of AI’s recent acceleration. Yet coming from someone who walked away from a successful physics career to work on AI, the prediction demands serious consideration.
If Kaplan is even partially correct, the implications extend far beyond physics. The question isn’t whether AI will transform scientific research, but how quickly—and whether our institutions, funding models, and educational systems can adapt in time.
For now, theoretical physicists have at least a couple of years. After that, according to one of their former colleagues, all bets are off.