University AI researchers are being squeezed out of their own field
The cutting edge of AI has moved from campuses to private labs. Academics at a recent Schmidt Sciences gathering described a field that's harder to fund, harder to study, and harder to justify than before.
Key points
- Academic researchers can't access the internal workings of frontier AI models like ChatGPT or Claude, which are kept private by their corporate owners.
- Universities can't afford the specialised computer chips needed to train frontier AI models, according to researchers at the 2025 Schmidt Sciences AI2050 convening in Mountain View, California.
- A Johns Hopkins study found that AI language models give less detailed answers to questions phrased in ways more commonly associated with women than men.
- Google DeepMind's AlphaFold team, which won a Nobel Prize for predicting how proteins fold, was disbanded last month.
- Some researchers worry that AI tools solving advanced mathematics could threaten the future of human mathematicians.
Picture being a marine biologist whose ocean sits entirely behind a corporate paywall. That's roughly how Nika Haghtalab, a computer science professor at UC Berkeley, describes academic AI research right now. She told MIT Technology Review that studying today's AI resembles being a biologist locked out of CRISPR, the gene-editing technology that transformed medicine, because private companies held it exclusively.
That comparison landed at a hotel in Mountain View, California, where the Schmidt Sciences AI2050 programme, funded by Eric and Wendy Schmidt, brought together some of the most respected AI academics working outside big tech. We covered this programme's research agenda on 10 August 2026, when the question of what academic AI can do that frontier labs won't was already pressing.
Why can't universities just do the research themselves?
Money. Training or running a frontier AI model requires enormous numbers of GPUs, the specialised computer chips that handle AI's heavy computation, and universities can't afford enough of them. The AI2050 programme gives fellows some funding to buy chips, which researchers called a genuine lifeline, but federal science funding cuts are tightening budgets further.
Even researchers who don't train their own models face costs. Repeatedly querying OpenAI's or Anthropic's systems to study them properly adds up fast, and neither company opens its internal architecture to outside scrutiny.
So academics are staking out different ground. Anjalie Field, a computer science professor at Johns Hopkins, put it plainly: "I try not to work on problems that I think are gonna be solved by a tech company." Her recent study found that AI language models give less detailed answers to prompts written in styles more commonly used by women. That kind of result, which reflects poorly on a company's product, won't come from inside one.
What about AI research that has nothing to do with chatbots?
A large slice of academic AI has never touched large language models. These researchers build focused tools: models that predict extreme weather or simulate how drugs interact with the body. They're not chasing OpenAI's goals.
But they've got their own problem. When most people hear "AI" they picture energy-hungry chatbots, and that image makes it harder for climate or health researchers to argue for their work's value. The disbanding of Google DeepMind's AlphaFold team last month, despite its Nobel Prize-winning work on protein-structure prediction, underlines how quickly priorities shift even inside the labs.
The field's talent pipeline is shifting too. Several well-known academics have taken leave from universities to join frontier labs, and many AI2050 fellows now hold academic posts alongside industry roles.
One newer pressure: OpenAI's models have solved real research problems in mathematics over the past six months, and some researchers are worried that pure mathematics may not have a long-term role for humans. One fellow told MIT Technology Review she was concerned about her mathematician peers' mental health.
There's a brighter note. Resource constraints push academic labs to make AI models smaller and faster to run. Tim Dettmers, a computer scientist at Carnegie Mellon who works on efficiency, argues that AI tools could make human scientists dramatically more productive rather than redundant. Constraints breed creativity, and always have.
If the next real breakthrough comes from a cash-strapped university lab rather than a trillion-dollar company, nobody in that Mountain View hotel room will be shocked.
