Universities' AI Talent Poached by Tech Giants - Wall Street Journal
[ad_1]
Alphabet Inc.’s Google division last week hired the director of Stanford University’s artificial intelligence lab to lead a new AI unit, the latest in a long line of academic stars in artificial intelligence lured away by tech giants.
Fei-Fei Li, a respected computer scientist, wrote in a Facebook post that she joined Google partly to “democratize AI.” She joined several other top professors who have left academia in recent years for tech industry posts.
In recent years, Geoffrey Hinton joined Google from the University of Toronto, New York University’s Yann LeCun went to Facebook Inc., Stanford’s Andrew Ng joined Baidu Inc., and Carnegie Mellon University’s Alex Smola moved to Amazon Inc. Many of these researchers continue to work at the universities in a diminished capacity.
Researchers warn that tech companies are draining universities of the scientists responsible for cultivating the next generation of researchers and who contribute to solving pressing problems in fields ranging from astronomy to environmental science to physics.
The share of newly minted U.S. computer-science Ph.D.s taking industry jobs has risen to 57% from 38% over the last decade, according to data from the National Science Foundation. Though the number of Ph.D.s in the field has grown, the proportion staying in academia has hit “a historic low,” according to the Computing Research Association, an industry group.
Such moves could have a long-term impact on the number of graduates available for teaching positions because it takes three to five years to earn a doctorate in computer science. “People are starting to question whether we are, in some sense, jeopardizing our ability to meet industry demand in the future,” said Mark Riedl, the director of the Georgia Institute of Technology’s Entertainment Intelligence Lab.
The squeeze is especially tight in deep learning, an AI technique that has played a crucial role in moneymaking services like online image search, language translation and ad placement, said Yoshua Bengio, director of the University of Montreal’s Montreal Institute for Learning Algorithms. Students in artificial intelligence are “worth somewhere between $5 million and $10 million to a company’s bottom line,” said Andrew Moore, dean of CMU’s School of Computer Science.
The tech giants offer perks that few universities can match, including steady funding, immense data sets and computing power, and the thrill of creating products that touch millions of people. Corporate salaries are much higher and may include a potential bonanza in stock options. The median annual salary for university postdocs in computer and information sciences was $55,000 in 2014, compared with $110,000 in industry labs, according to NSF data.
Researchers worry that the brain drain will interfere with progress in academic efforts to solve serious problems in fields like environmental science. “I am concerned that it will slow down our [rate of] discovery in the university and research labs because some of the best and brightest won’t be here,” said Sue Haupt of the National Center for Atmospheric Research. One of her top researchers was hired by Weather Company, now a division of International Business Machines Corp.
Carnegie Mellon has responded to the talent war by allowing faculty to cycle between the university and industry. Dr. Moore assumes that between 10% and 20% of people he hires will take “leaves of absence at any given time” to work in industry or to found a startup. “We really are expecting people to rotate in and out.”
AI experts at tech giants including Microsoft, Google and Facebook say they are attempting to mitigate the exodus from academia to industry by funding university departments and training students. IBM recently launched Cognitive Horizons Network, a consortium of six schools that aims to keep academics in university positions, according to Guru Banavar, vice president for IBM Research. This week, Google pledged $3.4 million to Dr. Bengio’s AI lab at the University of Montreal, which is part of IBM’s consortium. Facebook, Baidu and Microsoft said they also fund academic research.
Apple and Amazon declined to comment.
In some cases, tech companies allow their academic hires to retain university appointments. But employment curtails time spent teaching and sometimes limits discussion of topics related to work. Stanford’s Dr. Li said in her Facebook post that she would continue teaching at Stanford some quarters, though she didn’t offer details. Dr. Hinton splits his time between Google and the University of Toronto, where he still mentors three graduate students, he said. Dr. LeCun, who heads Facebook AI Research, is teaching one class this academic year, down from two before his employment at Facebook, according to NYU’s course catalog.
“Although it is very good that a number of [industry] researchers often continue in some role at their academic institution, it is of course not quite the same for the students,” said Bart Selman, a Cornell University AI professor. Enrollment in AI-related classes at some schools has more than tripled recently, he and others said, leaving remaining faculty members with far less time per student.
—Deepa Seetharaman contributed to this article.
Write to Daniela Hernandez at daniela.hernandez@wsj.com and Rachael King at rachael.king@wsj.com
[ad_2]
No comments :