Facebook Says It’s Not Destroying Academia With An AI Brain Drain

Late last week, Facebook announced the opening of new AI labs in Pittsburgh and Seattle. But the company says most of its researchers maintain academic affiliations, and help train their successors.

Facebook Says It’s Not Destroying Academia With An AI Brain Drain

Late last week, Facebook announced the opening of new artificial intelligence labs in Seattle and Pittsburgh, bolstering a global team of more than 150 researchers spread across facilities in Silicon Valley, New York, Paris, Montreal, and Tel Aviv. But though the people leading the new labs are coming from universities, Facebook says it’s not contributing to the demise of AI research in academia.


University of Washington researcher Luke Zettlemoyer will head up the Seattle lab and keep his faculty position, while Carnegie Mellon researchers Abhinav Gupta and Jessica Hodgins will spearhead the Pittsburgh office and keep their university affiliation on a part-time basis.

Companies like Facebook, Microsoft, and others say the ability for AI researchers to continue their academic work has proven to be a valuable tool for recruiting the best talent. At the same time, large tech companies have learned that allowing their researchers to publish much of their work in peer-reviewed academic journals and speak at academic conferences is also key to those people agreeing to accept positions in the private sector.

In a blog post, Facebook AI Research (FAIR) head Yann Lecun noted that it’s a common practice for the company’s researchers to stay involved with universities. Lecun himself has a dual appointment, splitting his time between Facebook and New York University. Others who work on AI at Facebook–a vital research area that the company is employing to handle everything from image recognition to battling objectionable content, and which Mark Zuckerberg told Congress was a major element in solving big privacy and security issues–maintain affiliations with institutions like Tel Aviv University, U.C. Berkeley, McGill, the University of Montreal, Georgia Tech, University College London, and New York University.

“All of us teach classes, advise graduate students and postdoctoral researchers, and participate in the life of our academic departments,” Lecun wrote in his post. “Our time is split 80/20, 50/50 or 20/80. Additionally, full-time FAIR researchers have affiliations with universities that allow them to advise PhD students….Those of us who come from academia continue to educate the next generation of researchers and engineers.”

Lecun’s post was meant in part to dispute a New York Times story suggesting Facebook’s new AI labs, and massive salaries being paid to AI researchers across the industry, are cannibalizing academic AI institutions. Lecun, the inventor of convolutional neural networks, one of the most important areas of deep learning, said the article “erroneously qualifies this evolution as a ‘brain drain’ from academia.”

The Times wrote that the new Facebook hirings in Seattle and Pittsburgh add “pressure on universities and nonprofit A.I. research operations, which are already struggling to retain professors and other employees.” Further, the article suggests that the demand for AI talent isn’t keeping up with the number of researchers, leading to steeply rising salaries. “Well-known researchers are receiving compensation in salary, bonuses, and stock worth millions of dollars,” The Times wrote. “Many in the field worry that the talent drain from academia could have a lasting impact in the United States and other countries, simply because schools won’t have the teachers they need to educate the next generation of A.I. experts.”


The article quoted Ed Lazowska, chairman of the computer science and engineering department at the University of Washington, as saying he was concerned that the large internet companies were luring too many of the university’s professors into the commercial sector.

Lecun counters that “Facebook is careful not to deplete universities from their best faculty, by making it easy to maintain sizeable research and teaching activities in their academic labs. In fact, making these part-time splits possible is precisely the reason why we have been establishing labs in New York, Paris, Montréal, Tel Aviv, and now Seattle and Pittsburgh.”

‘AI in academia, in general, is kind of screwed’

Mark Johnson, the CEO of Descartes Labs, a startup using AI to analyze huge data sets of satellite imagery, agrees that big salaries are playing a role in researchers leaving universities, but adds that “AI in academia, in general, is kind of screwed” regardless.

In his former life, Johnson was an executive in Microsoft’s Bing division, and he remembers scoffing at academic research papers on CAPTCHAs that touted working with 100,000 or maybe a million URLs. “We’re like, ‘Oh, that’s great. What happens when you have 100 billion URLs,'” Johnson recalls, “‘because everything in this paper is not going to transfer to several orders of magnitude, 1,000 times more data.’ So, yeah, a lot of the great machine learning research has already been happening [and] a lot of the great AI research has already been happening in companies.”

Descartes’s CTO and co-founder, Mike Warren, takes it even further. He thinks the notion of even doing meaningful AI research at universities or national laboratories is out the window these days due to endless red tape. “Why would you be in a university in the first place,” Warren, a longtime veteran of the Los Alamos National Laboratory, says. “They’ve turned into a complete bureaucratic mess….So don’t even talk about the salaries. It’s just not the type of job it used to be. It’s the same at the national labs. I was [at LANL] for 25 years. It’s complete blasphemy for a scientist with that sort of seniority at the laboratory to just pack up and leave for a startup” with uncertain chances of success.

Eating the Seed Corn

Still, AI researchers don’t grow on trees–they need to be trained somewhere, and there are concerns that the continued hiring away of academic researchers–especially when they don’t retain their university affiliation–will damage the reputations and teaching abilities of some of the most prestigious schools.


For example, The Times suggested that the opening of Facebook’s Pittsburgh lab will harm Carnegie Mellon, which lost 40 researchers and engineers to Uber in 2015, and recently lost another top talent to JPMorgan Chase. “It is worrisome that they are eating the seed corn,” Dan Weld, a University of Washington computer science professor told the Times. “If we lose all our faculty, it will be hard to keep preparing the next generation of researchers.”

Lecun was not available to comment for this article, but speaking to Fast Company in 2016, he specifically touted Facebook’s commitment to a philosophy of openness as a way of recruiting top AI talent–while also allowing many of them to remain connected to their research institutions.

Still, says Johnson, many AI-focused tech companies are keenly aware that they have little choice but to let their researchers be as open as possible. “We encourage people to go to academic conferences,” Johnson says, “because they write a paper and other people see all the cool work they’re doing here and realize that not all great science happens at universities [or national] labs. Great science happens at startups.”

And besides, he adds, the community of AI and machine learning researchers is tight, regardless of where they work. “People think that Google, Microsoft, and Facebook are hoarding all these AI researchers, and not [telling] anybody what they do,” Johnson says. “But these people move around from company to company. They’re all buddies. All the [Carnegie Mellon] people still hang out together, and they share what’s going on. And most [academic] papers are crap anyway.”


About the author

Daniel Terdiman is a San Francisco-based technology journalist with nearly 20 years of experience. A veteran of CNET and VentureBeat, Daniel has also written for Wired, The New York Times, Time, and many other publications