Anthropic has recruited such a high caliber of professors that it has become a pillar in the profession. “‘I’m joining Anthropic’ is the new meme right now,” Subbarao Kambhampati, a computer science professor at Arizona State University (who has not joined Anthropic), told us. This month, an AI company hired the chair of UC Berkeley’s department of electrical engineering and computer science, presumably to help build more capable robots. Perhaps most surprisingly, Anthropic in recent weeks has also picked up a Stanford economist, a theoretical physicist from the University of Maryland, and an analytic philosopher from UT Austin.
AI companies are turning out to be something like mini-universities in their own right. OpenAI employs leading mathematicians and physicists—including one who studies black holes, and another who specializes in string theory. At least three computer science professors joined Meta’s AI lab in late June. DeepMind, like Anthropic, is home to a a group of philosophers. And Anthropic’s recent publications indicate a willingness to hire legal scholars and political scientists. It is not known how many current and former professors are working in AI companies. In these four companies, we found more than 80—most of whom are computer scientists. Some have dropped out entirely; others still work part-time at the university. The number is probably very low, because we do not have access to local data; it also excludes many professors who have started their own companies, those at other AI centers, and those who work with industry in an informal capacity.
Especially for AI researchers, these companies are very attractive. “A lot of important research is being done in the industry now,” Humphrey Shi, a professor of computer science at Georgia Tech who joined Nvidia as a vice president last fall, told us. As he observes, “If you want to do something that’s really important, you probably want to join one of those bodies.” Tech companies are making offers—including very attractive salaries—that are hard for academics to refuse. In the process, research that would previously have been held in public is closed behind closed doors.
For decades, universities were the center of AI research. The field itself officially began at a research gathering at Dartmouth in 1956, and federal funding provided early support for the field. In the early 2010s, Silicon Valley executives began to take the commercial potential of AI more seriously, and set out to hire the best researchers. In 2013, Google paid $44 million to acquire a startup company run by three AI researchers from the University of Toronto. Facebook then hired Yann LeCun, an NYU professor, to set up the company’s AI research lab; Uber poached 40 Carnegie Mellon researchers to work on driverless cars. But for the most part, even as more work was being done within private companies, many of the new academics hired at tech companies kept their professorships and established an open research culture. “Researchers will be strongly encouraged to publish their work,” OpenAI wrote in its founding announcement. (Note the name of the organization.) Usually this helped lead to the current AI boom: In 2017, scientists at Google published a research paper that was. immediately of interest to OpenAI. Google’s invention, called “transformer,” is T in ChatGPT manages.
As AI has taken center stage, Silicon Valley has stepped up its efforts to recruit star researchers—and to look beyond computer science departments. Philosophers help train tech companies’ robots to better interact with humans, and economists examine the labor market effects of AI. Compensation is only part of the picture. The current era of AI research requires enormous amounts of computing power. Universities have only a fraction of the resources that Silicon Valley can provide, and the gap has widened as the Trump administration has cut scientific funding. Anca Dragan, a UC Berkeley computer scientist who leads DeepMind’s AI security and mediation department, wrote to us that she was motivated in part to take on the task to get “data, compute, and budget access to make progress on border security.” Some professors who remain in academia are also forming partnerships with frontier labs or starting their own AI companies, in part so they can pursue their research without resource constraints. And in some cases, he would star Ph.D. students are abandoning their graduate programs—or skipping them altogether—to pursue careers in AI instead.
Although most research still takes place inside universities, many professors told us, the result is a flywheel: As more academics are attracted to industry, the center of AI research is moving further away from academia, increasing the incentive for remaining researchers to leave. Moving can have some benefits. Many professors are currently in the lab on temporary leave; others may return to the profession full time. In doing so, they will bring new knowledge about frontier research to their institutions. (Dragan is currently preparing to teach a small Ph.D. seminar on AI security.) Scientists in AI labs can also innovate at a faster pace. After all, they don’t have to worry about applying for grants or waiting years for articles to go through peer review. “If some of the best research is being done at these companies, then I want my colleagues and faculty members to be there,” Chris Gregg, a Stanford computer scientist, told us. “You go where the best research is being done, and if that happens in a company, so be it.”
But as academics move away from AI labs or spend more of their time working at their own companies, universities are left with fewer professors to teach the next generation of researchers. (Gregg said students often ask why courses are no longer offered. The answer is sometimes that the professor is on vacation at an AI company.) When star professors leave universities, it becomes harder for students to do jobs that help them stand out to potential employers, Shi said. In order to differentiate yourself from other applicants, it helps to work with consultants who conduct research of results or cutting edge. Jennifer Chayes, dean of the College of Computing, Data Science, and Society at UC Berkeley, told us she fears that as research becomes confined to labs and proprietary models, it will be difficult for people outside the lab to use AI models to advance science. “Computer science departments in universities will benefit from this,” he said. “I don’t know if our innovation economy will.”
As the AI race heats up, companies are now refusing to release much of their research for fear of giving away their competitive advantage. Gregg and Chayes said they hear from researchers who have gone to AI labs that they can’t publish the work they want. “A few top AI companies share some things that are on the cutting edge, but it’s a very narrow slice of the AI literature that exists,” Nathan Lambert, an independent AI researcher, told us. According to one studywhile AI experts are making the transition from universities to companies, they are publishing about 65 percent fewer papers per year. The papers that are published tend to emphasize progress on security efforts and not on AI’s frontier capabilities; Corporate research can help fuel the Hype market. With the competition between AI companies increasing, it’s not just research that’s closing in, but AI models as well. Last month, when Anthropic released a powerful new model called the Fable, the company announced that it would be invisible destroy it the model’s ability to perform some AI research. The move was billed as a security decision, but the outcry from the elite was huge. (Anthropic apologized and changed the policy.)
Elites within tech companies are purer. Shi, a Georgia Tech professor who works at Nvidia, said the company is committed to open research, which is why he took the job. Meanwhile, Dragan, a Berkeley computer scientist at DeepMind, explained that printing is not everything. “I see a greater priority than publishing is to engage in policy and regulation,” he wrote to us. We also reached out to spokespeople at OpenAI, Meta, and Anthropic for comment. Anthropic declined to comment; OpenAI and Meta did not respond.
Its implications could be huge, not just for AI research, but for science more generally. In Silicon Valley, there is a popular view that the leading AI labs will end up as the primary drivers of science. “AGI has the potential to be the ultimate tool for advancing science and medicine,” said DeepMind CEO Demis Hassabis. he wrote last week. Hassabis before he said that the inspiration for his company was Bell Labs, AT&T’s R&D organization, which employed some 1,200 Ph.Ds at its height in the 1960s. Scientists at the lab won at least 10 Nobel Prizes, and invented the transistor and the modern solar cell. Already, DeepMind has conducted Nobel Prize-winning research on protein structure prediction, and advanced models have already shown. attractive ability in mathematical research. On Sunday, a mathematician working at Anthropic has been published that he had used the example of the Fable company during the World Cup final to solve a mathematical problem that lasted for about 90 years.
But if scientific talent and computing power are concentrated within private companies, Silicon Valley may also end up as the gatekeeper of science. Some are raising concerns: A group of scholars recently published a declaration warning about the “increasing involvement of technology companies in mathematical research.” Left unchecked, they said, technology companies’ involvement in research could affect “the scope and depth of mathematical research itself.” Rather than speeding up science, as AI company leaders claim it will, they may end up slowing it down.
