Facebook regularly generates a lot of media attention, much of it bad (between data breaches, Russian bots and questionable practices regarding data privacy, it is hard to find something good to say about that company) but this past summer it found itself at the center of a small controversy in academia when it announced its dual affiliation model, where its computer science faculty in artificial intelligence working at its labs in Seattle and Pittsburgh would commit 80% of their time to their industry jobs and 20% to their academic position, in a move designed to boost its AI labs by hiring from academia. (And it says other Silicon Valley companies should work more closely with academia on AI too.) The 20% devoted to “other pursuits” is reminiscent of Google allowing their staff to work on their own endeavors 20% of their time, or academia allowing its professors to work on consulting projects one day of the week, but a key difference is that Facebook specifically allowed its new hires to remain involved in their previous workplaces. The other option for Facebook would have been to have those academics sever their ties with their academic institutions. Given the scale of Facebook’s hiring, this could have raised some serious PR issues. (Amazon has also hired extensively from faculty ranks and other companies, with varying success. I’m not aware of a dual affiliation model at Amazon. I suppose that says something about Amazon’s demands on its employees’ time.)
Facebook’s dual affiliation model, although undoubtedly intended to assuage potential critics, has still raised concern among various academics, as universities in Seattle and Pittsburgh have seen some of their AI professors leave for the greener pastures of industry. The words “poaching” and “brain drain” have been bandied about. Yet, anyone in academia knows you don’t do that job for the money. Put another way, engineering academics’ primary driver isn’t their salaries. They wouldn’t be in academia if it were. Why some of them have considered industry after a successful career in academia points at some of the dissatisfactions that come with an academic job for superstar engineering performers, which Facebook, and other giants like Google or Amazon, have smartly tapped into.
Yet, the only surprise may be that it took so long for the problem to take center stage. As more universities seek to become world-renowned research universities, top engineering PhDs who might have otherwise gone to industry enter the ranks of academia and face the pressure of writing ever more grant proposals to fund ever more demanding universities and advising students who, unless the PhDs were hired at a university similar to the one they graduated from, may sometimes fall way short of their advisors’ expectations. Some of the CS/AI PhD programs at the universities Facebook has been hiring from may not quite be of the caliber of the institutions the “poachees” graduated from, and that can wear some people down. On the other hand, I remember reading a post from a former Harvard Assistant Professor who went to work full-time at a research lab, perhaps Google if my memory is correct, and there is no doubt that the Harvard undergrads are top caliber. Which brings us to the following point.
Engineering academia is a difficult path for many reasons, one of which being the expectation to train graduate students as soon as one begins on the tenure track, which requires a shift from producer to manager that employees in industry will only face after a few years as individual performer. Universities hire for tenure-track positions PhD grads who have done excellent work (and by the time they graduated they were doing that job basically by themselves with only minimal supervision from their advisor), and then those universities want said PhD grads to derive their fulfillment by training others instead of doing the work themselves, through indirect contribution rather than direct contribution. This can be very exhilarating for the right advisor-student pair but it can be tempting for a professor to, every so often, miss the pride that comes from making the primary, direct contribution in solving a complex engineering problem, especially in difficult university environments. (And this can also be true of business academia, where there is no pressure to get grants but professors must spend more time keeping MBA students happy. A PhD grad I know who was on the faculty at a great business school also left to work a top tech company. You don’t see the reverse happen nearly as often: industry employees rarely apply for tenure-track jobs, although some become Professors of Practice later.)
But is Facebook’s model the right way to envision academia-industry affiliation? Academia offers sabbaticals, which can allow faculty to get a taste of industry, yet industry projects don’t necessarily fit neatly in an academic’s calendar, and maintaining a foot in industry can help professors keep their skills up to date. This can benefit students in the classroom. A model where industry work and academic work are more entwined time-wise instead of being kept separate (“this semester I’m on leave, next semester I’ll be back with a full teaching load“) can also help faculty make more continuous use of all their strengths, especially since sabbaticals are few and far between.
Yet, there is no doubt that wearing multiple “hats” can lead to scattered energy and make it much more difficult to evaluate performance accurately. The very 80/20 time split, with 80% going to industry, suggests that Facebook recognizes as well as academics the perils of trying to serve two masters, and the split puts in no uncertain terms that the main master is Facebook. (According to a New York Times article, Facebook pays those professors handsomely.)
What problem was Facebook trying to solve for itself when it came up with this arrangement? There is no lack of entry-level PhDs eager to work in AI, but finding good technical managers with the cutting-edge scientific expertise required, whether to work at Facebook or Amazon or the other tech behemoths, can be a more arduous task. Young engineering PhDs haven’t succeeded thanks to their collaborative spirit. (Writing a dissertation is inherently a lonely endeavor.) Few technical managers at other companies have faced problems requiring the scientific expertise that Facebook needs. But the best engineering profs have succeeded thanks to their demonstrated ability in leading a technical team to solve cutting-edge problems. In that context, Facebook’s dual affiliation model was a masterstroke. Unfortunately, Facebook is plagued by such poor leadership by MZ – and perhaps SS, although I like SS – and so much bad PR (rightfully so) that the company itself runs the danger of scaring great performers away.
Instead of a 80/20 time split, perhaps a 20/80 time split, where professors would be brought in as consultants, or even a 40/60 time split, might prove to be a better arrangement for Facebook and the universities it is hiring from. Of course this would greatly reduce the salary that professors would earn from Facebook, but then, they can also quit and go to Facebook outright if that is more appealing to them.
It would also be good for Facebook to hire technical managers who can help it address its image problem, and if it positioned its AI hires in that context, that could certainly change its perception of a has-been company that tries to be too much to too many people in the relentless quest to have more users. Facebook is at risk of being supplanted by other social-media platforms, including Instagram, which it bought in 2012 for $1bn. (Instagram’s co-founders recently resigned from Facebook, and FB’s data breach also appears to have compromised Instagram accounts. Instagram may be Facebook’s best hope, but Facebook might ruin it first.)
As Facebook, Amazon and others try to poach professors from local universities, maybe more universities will come up with creative arrangements to keep their faculty members (more than one day a week). But we all have to respect the yearning of those professors who take the leap toward industry, whether mostly full time or not, to have a direct, quantifiable impact on a problem a well-known company faces. If this is the way they feel they will use their skills to the fullest, so be it. Fewer barriers between academia and industry can only increase satisfaction for the professors tempted to explore both worlds. I’m not one of them, but I don’t think other professors or university administrators should tell those professors who wandered through Facebook’s doors how to live their life. (It smacks of resentment and self-interest.) If those professors want to stay at Facebook, good for them. And if they ultimately prefer returning to their university work full-time, the doors of academia should be wide open for them.