Facebook Adds A.I. Labs in Seattle and Pittsburgh, Pressuring Local Universities | Hacker News Reader

Facebook Adds A.I. Labs in Seattle and Pittsburgh, Pressuring Local Universities

nytimes.com

67 pointsGW15091441 comments

> “It is worrisome that they are eating the seed corn...If we lose all our faculty, it will be hard to keep preparing the next generation of researchers.”

It is interesting to note that in our free-market(ish) society, preparing the next generation of researchers is not as valued as working at a BigCo or a Unicorn. If AI professors are in high demand relative to the supple, basic economics dictate that their prices should spike up. Then why aren't these professors getting paid a lot more to continue teaching?

There is usually a cap on pay for professors. Maybe that is factoring in?

Alternatively, universities could act as startup incubators, a la SRI (Stanford Research Institute, who created the prototype for Siri.)

Professors are being paid below market value for their skills. If tenure came with a replenishing bucket of venture capital, under stipulation of open source/patent sharing, it'd give them real money, autonomy and great internships for their students!

Schools would make bank, too.

Or, alternatively, schools could actually act like academic institutions and quit wasting exorbitant amounts of money on bullshit like sports and unnecessary construction projects which drive low-quality student enrollment.

You're making common mistakes that people make when talking about school finances. Schools pay for sports because they drive alumni interaction and therefore donation rates, and they therefore have a higher than 1-1 rate of return. They have the unnecessary construction projects because the funds are earmarked for said construction. If I give you, a school, $X million, on the condition that you build a new student hall with my name on it, will you turn that down?

If I offer you $1200, on the condition that you spend that money on a 36 hour trip to Hong Kong and back, will you turn that down?

I assume so. So there has to be some other force at play, unless the university admins are just maximizing revenue in a brain-dead fashion.

The analogy isn't a good one, as you get no value out of the trip to Hong Kong in your scenario.

A more relevant analogy is if I offered you $1,200 on the condition you spend that money on a home gym. Could you have other uses for the money? Sure. Could you have higher priorities than exercise? Sure.

Are you going to turn that down? Probably not.

An even more appropriate analogy is if I offered you $2,500, but $1,200 had to be spent on a home gym. The remainder has no conditions. That would be closer to what a lot of big donations are structured as.

You're leaving out aspects of that analogy. You left out the part where you incur the added cost of electricity, or your current home isn't big enough so you need to move to a new area with higher property taxes and insurance costs. Or maybe you're renting the home to someone and force them to get a new gym in exchange for higher rent. The point is, who actually benefits from the donation; the fancy pants administrators and donators or the actual students who'd probably prefer lower tuition and better quality instruction? Somehow, big donations never seem to translate to the the obvious benefit of affordability and accomplishing the actual goal of providing a good education to people.

If it is a taxpayer funded institution, then it should be turned down. Donations should be anonymous, otherwise I assume the donor is wanting and getting something in return (nepotism, brand name, hubris), and that doesn't belong in taxpayer funded arenas, even street names and parks and whatnot.

I don’t think anyone wants to receive hubris in return. Not that a university is in a position to offer it.

There's a reason that people want their name on a donation. If it was truly selfless, it would be anonymous. Here's an example:

https://www.nytimes.com/2018/05/05/style/stephen-schwarzman-...

I guess I don't get your point? So what? Really not trying to be rude, just curious why this matters.

Oh I see your post earlier - yeah, I think removing incentives from voluntary contributions to (partially) taxpayer funded events would lead to a major increase in inefficiency of those projects seeing the light of day.

> Then why aren't these professors getting paid a lot more to continue teaching?

Because the majority of money that a professor controls mostly comes from grants. And you have to fight for grants from an ever shrinking base.

Why fight for grants when you can get paid a dramatically higher salary and can focus on research instead of fundraising?

Until academia figures out how to fix its funding model, this is the new normal--any hot tech area is going to result in a brain drain.

Correct.

Professors bring in money by teaching, for which they aren't paid any extra, and grants, which often they receive a small monetary incentive, i.e., summer salary of about $10-20K per year for a professor making $100K. It is hard to get much more than that because of caps on grants, so for professors to get paid more by grants the maximum size of a grant from NSF/DoD/etc. would need to be a lot bigger.

A question I have is "Why aren't professors who teach larger and more popular classes paid more for taking on the additional students?" My deep learning classes fill up to the maximum size allowed (50). I could scale it even higher, although finding class space would be hard, but if I do that my university has no explicit mechanism to reward me for doing that. If you teach more people, you would think that you would make more money.

Too far removed from the day-to-day to be reflected that it is more important to keep teaching than go to big company. Likewise, research could never make it out of private companies and that is certainly a less favorable outcome for the public as a whole. The AI professors are probably paid much higher than peers in non-technical subjects but no university is going to be able to match Google or Facebook salaries, especially when these guys are in such demand that private industry is willing to pay $300-500k+ each for them (sometimes much more).

In general, AI professors are not paid any better by universities. Maybe if you were a superstar and had threatened to leave, but even then your compensation would be relatively small in comparison.

> Then why aren't these professors getting paid a lot more to continue teaching?

Paid with what money? Tuition is probably already set to the highest amount that is competitive with other schools.

So students pay insane tuitions and professors are relatively underpaid. Where does all the money go?

At CMU, annual tuition is less than $24,000. There are less than 14,000 students enrolled, which means total tuition paid must be less than $336M.

If all that money were equally divided between the 1400 faculty members, they'd each get a salary of at most $240k per year, which is really high in the grand scheme of things but kind of low for these specific facebook jobs that we're talking about.

Now consider that much (most?) of that revenue must actually be spent on the 2,500 staff members supporting university operations, building maintenence, etc.

I don't think there is any great mystery to solve here.

Edit: see the replies to this post. It seems the tuition at CMU I'd double what I thought it was. I still don't think the revised numbers are any more mysterious, though.

Undergraduate tuition at CMU is $52,732 per year

https://www.cmu.edu/news/stories/archives/2017/january/tuiti...

Indeed you are right! The collegedata page that Google took me to must have been woefully out of date.

Still, a mere doubling of the faculty salaries I was describing above is likely still not sufficient to compete with these tech giants when hiring someone like a deep learning expert.

One faculty member for ten students is also incredibly high.

So if someone if getting trained for a type of job which will generate millions of future earning potential for them, should the cost be only 24k (or even 48k) per year?

It really does seem kind of insane to stay in academia right now in AI. Far higher salaries are available at the moment in industry, although I expect that to go down over the next 10 years due to the flood of students and to tools becoming easier to use. You get paid maybe 25% of your worth in industry, spend a lot of time dealing with disgruntled/cheating/unhappy students, have to deal with academic meetings, and then spend most of your remaining time writing grants that have a 10% chance to succeed. As a professor, I work about 60-80 hours per week, with about 30 hours being for meetings, teaching, etc. The remaining time is spend editing, reading, and writing (grants/papers).

On the plus side, you get to run your own lab to work on the things you want (as long as you get money), be your own boss (mostly), and get the privilege of training the next generation of researchers and practitioners. I really feel like I have a lot of positive impact on people and my community as a professor, and I doubt I would have nearly as much individual impact if I worked in industry.

Meanwhile, Professors in the humanities departments at the University of Washington worry about how they'll afford to pay rent with Seattle's skyrocketing housing costs and the University's lack of raises for folks in their fields.

(source: I know several.)

But what is the solution? Raise taxes for all of us to pay the increased salaries of humanities professors?

There is certainly great value in humanities, but that value shouldn't be determined by "an unrelated field is hot, so let's make hot money too".

Especially when it's other people footing the bill, and especially with what's going on in humanity departments these days.

I don't think that being able to afford a place to live and being able to pay your student loans is "hot money."

Depending on your student loans and rent prices, wouldn't it be?

Rent in a major city is $2000/mo. Student loan payments can be in the tens of thousands per year. Assuming a normal tax rate, professors would need to make 6+ figures to live comfortably.

Step number zero should be to let housing get cheaper by relaxing zoning laws in the city.

It's almost like the market doesn't want to pay more for humanities professors.

That's not necessarily a bad thing.

It's almost as if when all you're able to do is compare two numbers and see which one is bigger, every problem gets reduced to comparing two numbers to see which one is bigger.

Eventually it might come down to entropy

Hmm, hire more faculty? Plenty of PhD students are waiting to get an academic job.

Every story about facebook now portrays them as satan.

I don't even like facebook but this is definitely a trend.

True, I mean at least since a year or so. I think last year articles about studies showing Facebook having an adverse effect on happyness arrived in the mainstream, i.e. Google News.

But anyways, the more I think about Facebook, the more question it raises. What I was wondering recently: how do they acutally pay the low-income personell, like guards or janitors? If you imagine that even people earning a 6 digit salary have a tough time finding an apartment and commuting to work - both subject to protests by locals - how do these people live?

I was thinking about that as I read a post on HN about exactly that problem at Yelp. The article was deleted though. Here's the reference: https://news.ycombinator.com/item?id=16844410 This makes me think there is something deeply wrong in the whole Startup bubble.

> how do they acutally pay the low-income personell, like guards or janitors?

They are most likely contractors. And they probably live far enough where they can afford living. But that means longer commute.

I've noticed this trend of trends. It's as if media outlets have become an echo chamber.

The media became incensed that Zuckerberg, till-then held as a paragon of progressivism, (indirectly) provided the same service to Trump that he did to Obama during the two previous election cycles. Suddenly the supposed idealist turned out to be a profit-first capitalist, willing to equally trade with left, right, and center. Of course it doesn't help that Facebook has been long known for playing shady tricks with things like user privacy.

Screw it, I’ll go against the grain. This is why it’s foolish for CS departments to invest so heavily into trendy areas - either the trend dies, and the faculty member is useless, or the trend is so hot that the faculty members will be hired away. So many schools are desperately hiring ML people to start their own data science stream, and they’re competing with major companies for the top talent.

But why do you say they invested particularly heavily into this area?

Response from Yann LeCun (https://www.facebook.com/yann.lecun/posts/10155263530987143)

Facebook AI Research is opening two new labs in Seattle and Pittsburgh, which will join the existing sites in Menlo Park, New York, Paris, Montréal, and Tel Aviv.

We are delighted to welcome Luke Zettlemoyer to the Seattle lab and Abhinav Gupta and Jessica Hodgins to the Pittsburgh lab. Luke will retain his faculty position at the University of Washington, and Abhinav and Jessica their positions at Carnegie Mellon University on a part-time basis.

Many FAIR researchers share their time between FAIR and a university: Lior Wolf (Tel Aviv University), Jitendra Malik (UC Berkeley), Joelle Pineau (McGill), Pascal Vincent (U Montréal), Devi Parikh and Dhruv Batra (Georgia Tech), Iasonas Kokkinos (UCL), Rob Fergus, Kyunghyun Cho and myself (NYU). All of us teach classes, advise graduate students and postdoctoral researchers, and participate in the life of our academic departments. 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. Examples include Leon Bottou and Jason Weston who are both affiliated with NYU. Our Paris lab hosts over 15 resident PhD students who are co-advised by a FAIR researcher and an academic advisor. Those of us who come from academia continue to educate the next generation of researchers and engineers.

This is made possible by the fact that we practice open research at FAIR, which makes it easy to collaborate with academic groups. This new modus operandi is redefining the relationship between academic research and industry research. Professors gain a different type of experience in industry that can have a positive impact on their students and on their research. Additionally, their connection with industry helps produce new scientific advances that may be difficult to achieve in an academic environment, and helps turn those advances into practical technology. Universities are familiar with the concept of faculty with part-time appointments in industry. It is common in medicine, law and business.

This NYT piece by Cade Metz erroneously qualifies this evolution as a "brain drain" from academia. But 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. It is the proximity to leading universities with talented faculty, and the existence of a local talent pool that attract us. Unlike others, we work with universities to find suitable arrangements and do not hire away large numbers of faculty into full-time positions bottled up behind a wall of non-disclosure agreements. We contribute to local ecosystem.

Faculty who join FAIR part-time practice open research, teach courses, advise PhD students, and participate in the life of their department.

Reply on news.ycombinator.com