Academia used to offer smart people the promise of tenure and interesting work. Now they offer eternal postdocs and publish or perish. Is it any wonder they can't retain good people?
This is talking mostly about perceived "stars" in academia, who already do have tenured or tenure-track positions (or offers of them), but leave for industry instead. So I think it's a different issue than the one you're talking about.
For those people, one thing that's changed is that academia, in STEM at least, has grown into having a big management component. You're expected to run a significantly sized lab, and to continually bring in enough grants to pay for it. Professors at top universities spend a large amount of their time writing grants, hiring people, and networking with potential funding sources, and it's no longer really optional. There is a certain style of researcher who actually excels in this kind of position, managing a medium-sized research enterprise, and advancing a field by putting together good teams, providing high-level direction, and mentorship. It's like running a startup in some ways (obv. different in other ways). But not every researcher is good at or interested in that, and some would rather opt out and find a job that lets them focus more on their individual research.
Another thing that's changed is pay. Academic salaries in CS have gone up significantly, to the point where $100k+ is common, and $150k+ for big names is also getting common, whereas professor jobs were traditionally more lower/mid-middle-class than upper-middle-class (more like "$75k+" than "$150k+"). But salaries at tech companies have gone up even faster. If a big name in deep learning gets offered $300k base plus a big stock package, the university's retention counteroffer isn't likely to come very close.
Also, specifically in AI (which this article focuses on), data access and engineering/infrastructure support are other lures of industry. If you do machine translation, for example, moving to Google will solve many of your dataset problems. If you need a distributed cluster, there's expert IT staff working on it; if you do autonomous vehicles, there's expert engineering staff supporting your "demo platform". In academia you rarely can afford that kind of infrastructure or engineering support, so there's a lot more of grad students cobbling together DIY demos, and the postdoc hired mainly for his/her mathematics knowledge also moonlighting as the cluster admin.
I think your numbers are off by about a factor of 2 (underestimating). You can see the salaries for the UoC system at https://ucannualwage.ucop.edu/wage/. Note that most of the professorial salaries are labeled AY which means they are probably making 30% more annually. Salaries at private schools are even higher. Silicon Valley is not going to lure a "big name" away from a tenured job for south of a half-million a year.
I think academia has a long way to go to reduce costs. I'm willing to cut services that universities offer if that means lower costs.
Remove the nice dorms, cafeteria, and fitness centers. Remove the community outreach programs that isn't really the job of a university. And yes increase the number of introductory level classes taught by graduate students and post docs.
I'm not worried about people moving from academia to industry. My hope is as we increase the rate at which people finish high school and go to college, we will have enough people for industry and academia. It is tempting to say we need to offer better wages and benefits for teachers but we can't digress. Cost containment should be our first priority.
I think it is possible if we understand and agree with the compromises we will need to make.
"Even more strikingly, an analysis by a professor at California Polytechnic University, Pomona, found that, while the total number of full-time faculty members in the C.S.U. system grew from 11,614 to 12,019 between 1975 and 2008, the total number of administrators grew from 3,800 to 12,183 — a 221 percent increase."
The administrators aren't really the problem per se, though. They're there to address mandates from various levels of government. Colleges won't reduce the number of administrators because they can't under the current bureaucratic regime.
How do we know that you are correct? Maybe universities do have bloated staffs. Salaries, benefits, retirement are going to be a lot more costly than nicer dorms and a fitness center.
People are offering solutions without following the money.
"The compliance problem is exacerbated by the sheer volume of mandates—approximately 2,000 pages of text—and the reality that the Department of Education issues official guidance to amend or clarify its rules at a rate of more than one document per work day. As a result, colleges and universities find themselves enmeshed in a jungle of red tape, facing rules that are often confusing and difficult to comply with."
You don't think colleges would rather have the money than all these extra paper pushers?
That depends on how the people at the top see themselves
- "We want all responsibilities, Minister, if they mean extra staff and bigger budgets. It's the breadth of our responsibilities that makes us important, makes YOU important, Minister. When you see vast buildings, huge staff and massive budgets, what do you conclude?"
- "Bureaucracy?"
- "No, Minister, you conclude that at the summit there are men of great stature and dignity who hold the world in their hands and tread the earth like princes."
Wow, you were not kidding at all. 148 people with pay over a million dollars. We need a much lower ceiling on pay when it comes to public institutions that are supposed to be not for profit.
My thought is a cap on the multiplier on the minimum wage (for example, a 50 times multiplier at a minimum wage of $15 per hour would mean an annual cap of $15 * 2000 * 50 or $1.5M). Similarly, a private corporation could declare that it will cap its pay at 100 times multiplier of the minimum wage.
I think this is something the Regents of the University of California system could easily impose if they had the will to do so. Would a pay policy like that be illegal? Would it be considered conspiring or "fixing" if it applies to everyone at the organization?
You misread the page. The table of top salaries contains separate rows for each year between 2004 and 2014. I can't figure out how to show only a single year, but it would be about 25 people making over $1M/yr (since most of those rows are since 2010).
I don't see what being not-for-profit has to do with salary caps. But the whole reason that the head coach is paid so much is that the football team is being run as a for-profit enterprise, which is a seemingly rational thing for the university to do, yet I think it makes sense to restrict universities from distracting themselves with that sort of enterprise related to their purpose.
Remove the nice dorms, cafeteria, and fitness centers.
You know the dorms and dining halls at these schools were pretty much the same 20 years ago (before tuition costs started mushrooming as insanely as they have) as they are now, right? So perhaps one should look elsewhere as to where the rising costs have been coming from.
(Fitness centers being an optional expense at many schools, and something from a different animal than the subject of basic operational + living expense costs).
This is certainly not true at some universities. For instance, at the University of Michigan dorms have been renovated on an annual cycle (around one building a year) at an expense of millions of dollars for small renovations and tens of millions of dollars for larger ones. These larger upgrades often include fancy dining halls with lots of options, along with well-appointed interiors. Furthermore, the number of new buildings and construction projects (about 10-12 currently) at any given time is insane in my opinion, though some are funded by donors for specific uses so can't be reallocated.
>My hope is as we increase the rate at which people finish high school and go to college
Woah, this is the logic that is getting us in trouble!
66% of people already go to college. Of those who go, only 50% graduate. So...
1) Does "college" matter or does learning a marketable skill mater? If we had 100% college enrollment and completion but everyone was getting a degree in native American studies, would we somehow be better off?
2) At what level of enrollment does it become more cost effective to simply extend high school 2-4 more years and just consolidate it?
College in its present form is in trouble because the lions share of college attendees are; white women, taking non STEM degrees, taking on federal loans, and having a lower workforce participation rate at the end of it.
College is about learning, not about what a huge chunk of it has become; safe spaces, community outreach, nice dorms, cafeterias, dating, fitness centers and generally taking a hiatus from life. We really need to redefine college in its present form, it is no longer what it was 50 or 100 years ago.
> We really need to redefine college in its present form, it is no longer what it was 50 or 100 years ago.
The norm in Australia is to "take a break from life" for 2 years, living on campus, and then to live like an adult, buying groceries and cooking food for the final two years.
This is important. Most probably, the sports is proving a big factor that is killing the education in USA [1]. Excessive indulgence of US educational institutes (schools, colleges and universities) in sports has been a very big mistake made by the US society.
The vast majority of colleges lose money on college sports. And when I say "money", I mean millions of dollars that come directly out of students' pockets in higher fees. Here are the numbers: http://www.al.com/sports/index.ssf/2014/08/ncaa_study_finds_...
"The frantic spending race is playing out differently across the country. Higher coaches salaries, while common, are just part of an array of expenses soaring at athletic departments that fail to profit.
At the University of California Berkeley, the mortgage on athletics buildings went from $0 to $23.4 million in a decade. At the University of Wisconsin, annual maintenance and spending on facilities went from $10.5 million to $38.2 million. At Florida State, pay for athletics staffers — not including coaches — went from $7.7 million to $15.7 million. At other schools, rising costs for travel, severance pay, recruiting and other items combine to keep athletics in the red."
I don't see why you can't have other extra-curricular profit centers in school. Why don't universities run application development, art, design, & business management together? Make some products to sell.
Is that true? I remember hearing that the University of Kentucky's basketball program ran a net profit, but also that it was pretty unusual in that respect.
Yes, this is also true for the University of Michigan athletics department, which in a recent year spent around $145 million and earned $158 million in revenue.
Of course, it's not enough for something to return a profit; it needs to return more profit than an alternative use of the funds. Perhaps it could have made more money investing in postgrad startups?
My understanding is that the UK basketball program is self-sustaining, and additionally turns a profit for the university. It's not taking university funding to begin with.
(Also, much as I'm disdainful of university athletics, declaring that only the most profitable programs have a place on campus seems like a really amazingly horrible idea.)
Postdocs aren't that common in computer science. They exist, for sure, but most of the recently hired CS professors I know came straight from their PhDs.
Most of the newly hired CS professors I know (at some legit schools - CMU, Waterloo, USC, UT Austin, etc) came from postdocs. Could be a sub-field of CS thing?
Definitely a subfield thing in part. Highly theoretical CS PhDs (complexity theory, algorithms, etc - the folks who are mathematicians in computer scientist's clothing) almost always require postdocs, both because they tend to be less heavily recruited by industry, and because the vast majority of departments are shying away from theory in favor of things that bring in more grant money and industry partnerships.
Departments aren't stupid, though. They realize that grabbing star grad students with highly marketable skills in industry as they are walking out of their dissertation defense is probably the only shot they'll have at getting them. They will often "try out" industry or take a job at Google or MSR on the advice of their PI, noting that it's just as good as a postdoc but pays four or five times as much. Problem is, they then have to walk away from their $180k/year to take a faculty position that might not pay half that and could very well dead-end in a few years if the department leadership changes (or any of a dozen other precipitating factors) and they don't get tenure.
All that said, I personally know of tenure-track professors at two of the schools you mentioned, plus ones at quite a few other top-flight CS departments (Cornell, Toronto, Washington, Cal), who were hired straight out of grad school.
I think you have a bias of looking at students from select top schools. A PhD student in almost any branch is going to have a really hard time getting a tenure track position unless they have an awesome research record (publications in top tier conferences/journals).
both can be true. super rough approximation for CS faculty job candidate consideration: Ph.D. outside of the U.S. followed by a postdoc in U.S. at School X ~ Ph.D. in the U.S. at School X
In other disciplines (e.g. mine - theoretical physics) it is not possible (whereas doing postdocs for 10 years or so is common).
And while it is harder to transit (learning basic computer skills, while way easier than quantum field theory, takes time and effort), the opportunity gap is way higher. And it may be not a surprise that many data scientists with PhDs are physicists.
Even if true (I know many professors who did postdocs), it's not clear what you think this proves. Faculty slots are very scarce and most newly graduated PhDs, even very smart ones, do not have "faculty job at a desirable institution" as an option immediately available to them. So they are left with the choice between postdocs and industry. Most choose industry, and it's not hard to understand why.
Postdocs especially multi-year postdocs are very rare in Computer Science. I have personally known several students get hired directly or after a one year post-docs. Multi-year postdocs are common in fields (Biology in particular) where job prospects offered by industry are comparable in salary and future opportunities to what postdocs offer.
that's actually absolutely correct :) academia is best at doing things that are not directly marketable in the capitalist economy. in the early 1900s, a group of weirdos working on theoretical quantum mechanics couldn't find marketable jobs in industry of the day, so they had to do their work in academia. thanks to them, you have your iPhone now.
That's actually the crux of the issue here. In the past AI research didn't have any commercial value, and now it does. So universities and departments are having to deal with new competition, and they'll have to adjust various forms of compensation as a result.
The article talks about "stars", which implies quality over quantity. They probably already crank out too many infinitesimal-delta-thesis guys who are smart but not brilliant.
Can confirm almost all Computer Vision professors I know work or have worked at Google/Facebook/their own startup at some point of time. Also these Professors have tenures (currently on sabbatical etc.). This is actually a great thing for graduate students too. Internships pay very well and can even extend beyond summers as part-time contractor positions. Another common method as highlighted in the article is to convert research group into a startup which gets acqui-hired by Google etc. E.g. Geometric Intelligence would be a good example.
Maybe universities should consider paying more to compete with the tech companies. It's a free market and whatnot. If the price goes up, you gotta pay more. Complaining about the price going up isn't exactly productive.
I doubt universities can come close to affording it. My CV wouldn't have gotten me a tenure-track position at a top school, but my income in industry is much higher than even the best paid professors at my alma mater (ignoring profs' consulting income, of course).
To be honest I wonder if the academics that are still there just don't realize how lucrative industry is...
Then universities should charge even higher tuitions or shift some of the ridiculous money for tuitions they're already getting and burning into building over-the-top facilities around to offer these people more money, or just don't have a computer science program.
Over the past ten years I've seen my alma mater replace almost a third of its campus with brand new buildings 2 to 3 times the size of the old ones, and it has plans to replace almost half its campus by 2025 (its plan is a public document). The buildings it replaced were starting to show their age a bit, but they were still plenty usable (one building legitimately had health code violations. Ironically, it's one of the buildings that still hasn't been demolished yet).
Meanwhile tuition there has more than doubled in less than ten years. Some of that is due to the government reducing its subsidy, but a good chunk of it is going into these massive construction projects.
Some of that money can go to better salaries for professors getting lured by higher salaries elsewhere, surely (although I do worry more of it will go to administration salaries instead).
Whining about losing employees because you aren't willing to pay enough (or offer enough ancillary benefits to augment your pay), especially when your tuitions are astronomical anyway, just comes across as arrogant, and they deserve to lose their best people.
Tenure-track professors in computer science at top universities get paid about 100,000 to 250,000. You can look this up publicly because
a) California state employee salaries are online [1]
b) UC Berkeley is a top school for CS professors and competitive with everyone else on salaries. Looking at random AI faculty members shows about 100k at hiring, about 150k at tenure, while machine learning king of kings Michael I Jordan gets 300k.
Google pays starting undergrads > 160k, and fresh PhD students > 225k[2] (all in comp including first year vesting stock + base pay + expected 15% bonus). I imagine if they're poaching faculty, they're getting significantly more than that[3].
Basically I'd say the industry salaries are at least 2x the academic salaries, if not higher. The flagship hires are paid a lot more (think Yann LeCun, or Andrew Ng), and importantly, given the freedom to hire and build mini-empires to do stuff.
As a final note, I know talking about salaries publicly is awkward, but this is all public information, and grad students talk freely about this sort of stuff with each other, so I don't feel like I'm revealing anything super secret beyond showing you how to look up the relevant information in publicly available areas.
[2]: Source: Glassdoor Google for "research scientist", which is the position that PhDs in AI/ML are hired into.
[3]: Google levels are public, and while Glassdoor data is sparse, faculty who are poached are hired at at least the "senior staff engineer" level, which glassdoor says about 500k, which sounds around right to me, if not a little low. As an aside, salaries at the higher levels have much higher variance than at the lower levels, and to entice someone away from a tenured position probably requires hitting the higher ends of the salary band.
You're setting the bar for hiring faculty members far too high. Perhaps you mean tenured CS faculty? (Source: Stanford faculty in the process of moving to Google Research)
Brand-new CS assistant professor hires at any research-intensive CS department (there are ~80 of these schools in the U.S.) probably come in around $100k right now (as arjun said, many salaries are public) ... now that's a 9-month salary, so if you get grants to fund your summers (which you will if you're in the groove), that's a starting of $133k annual. That's not bad, considering most of these schools are located in low-ish cost of living areas. Of course, there are no stock options, huge bonuses, and other lavish benefits of industry, but it's definitely still a financially privileged position.
I meant that the parent's analysis in his footnote of what the minimum position and starting salary would be for a faculty member to ever move is setting the bar too high; it seems my wording was vague. His academic numbers seem spot on, but I question footnote [3].
Given that most academics have, almost by definition, previously prioritized their work environment over their salary, one shouldn't even assume that an academic receives an (immediate) multiple factor raise by making the transition.
225k starting?! Holy cow... so grad students that go on to work as a research scientist at Google literally increase their salary by an order of magnitude. That's crazy.
yep, but if you're good (e.g., a deep learning expert nowadays) and work on mission-critical, visible projects at a company with rising stock prices (Google, FB, etc.), i'm sure you can sustain $200k total comp each year as a fresh Ph.D. grad.
You want to really feel bad about yourself? I have a friend that makes $75k base + $75k in bonus each year. He also has stock that nets around $450k every 3 years. So all in comp per year he makes roughly $300,000. Did I mention he's 23 with no experience (BS degree to boot)?
I guess the shitty part is that he has to live in SV and work for Facebook. I'd be tempted to take an offer for $300k, but having to work at Facebook makes it almost not worth it. So I guess the comp matches the job.
It's just strange how much of gap there is for SEs and developers (or anyone in tech for that matter). You could be making $75k in the midwest working on something important and living pretty comfortably (if you have a spouse with income). Or you could be making $300k with less experience working at Facebook on some feature that'll never see the light of day. You're barely scraping by because the $2mm mortgage on your 1200 sqft house is killing you.
If it makes you feel any better, the rationalization I apply to those Google/FB salaries (total comp.) is that I'm fairly confident that I will never be able to work at either of those companies. And I am at peace with that since working in this technology industry bubble, allows me to work with my mind and with talented people with diverse interests and appreciate how fortunate I am to have the ability to have what I do have - not what I don't have.
They are trying to close the gap, but it's a huge gap. I was told by a professor at my Alma Mater that PhD students get a 40k stipend (in St. Louis). Contrast that to most other programs where it might be closer to 20k, and to the industry where an intern can easily make 80k per year, and a graduating senior qualified for grad school can pull 6 figures from many places.
For stars and talents, academia provides something the private sector can never provide: freedom.
Established professors can dictate their own rhythm. They can work hard all year long, or they can slack for some months and concentrate on their families more. They can change sub-fields of interest as they see fit, and work on problems they want to solve. Also, they get to teach which is by far the biggest privilege one could ever achieve.
Of course private research provides very interesting deals, but most stars and talents are exactly that because they had the freedom of doing things as they see fit. Putting them in a company structure with quirky rules is not their natural habitat.
In short, in academia you are bound by hierarchy, local politics, grants and bureaucracy, 1-2 year lag in anything. As a freelancer I have enough $$$ I am free to pursue my intellectual interests (any side projects, teaching students topics of my choosing, etc). And when it comes to teaching - I get _only_ students that are interested, not - diluted by the crowd who just wants to fulfill the curriculum.
I like your article, but certainly your PhD experience is not comparable to an established professor position which was really the hole point of this thread. PhD's are harsh different realities. You're still a student as a PhD and it shouldn't really be compared to anything else; certainly not a contractor position.
You're right that there are professors that are bound to hierarchy, local politics, grants, etc.. But the important point is that it is their choice to be bound to that, not someone else's.
David Graeber once said "Every society has something to do with brilliant, imaginative and extremely impractical people ...we used to put them in academia, but now academia is all about self marketing." - https://youtu.be/IHAJiuU5xhk?t=9s - I think this really hits the nail on the button. Academia was and should be a place where these impractical stars can shine to their potential.
I agree that the gap for academia to industry is very different for young and established researchers.
No, I am not a PhD student (I defended 1.5y ago). And I saw that I really don't want be a postdoc; not even as a "purgatory" to go through - I didn't envy professors either. While professors do have a lot to say, they are still bound by administrative duties, pointless meetings, maneuvering grant calls so they can use it for something they actually consider useful, etc.
Sure, for some its fine (or at least - acceptable; or - they don't have other intellectually challenging job opportunities). Some others (e.g. the ones in the article) are happy to join Google. And I am sure it is not only about money.
It's a very different experience than being a professor in 50s or 60s; e.g. one day I talked to (now late) Kenneth Wilson and he told me that he had changed his field overnight (into one that gave him the Nobel prize). Right now it is unimaginable.
Sure, I also consider the current shape of academia sad. But... when it comes to the "impractical" dreamers, let's not colorize the past. Vide my another writing: http://crastina.se/theres-no-projects-like-side-projects/ (for academia, look at the John Bell's example).
> They can work hard all year long, or they can slack for some months and concentrate on their families more
Yeah, that's kinda not true. In academia, you have the freedom to work whichever 60-80 hours a week you want, but no one becomes a "star" (or even an "associate professor") by slacking for months at a time.
> Also, they get to teach which is by far the biggest privilege one could ever achieve.
do you have any experience with academia? it is very much not free in spirit. even getting a ph.d. these days requires more exercise in tedium and logistics than creative thought.
Top talent has been fleeing academia for a while now. Lower pay, limited job security (unless you manage to win a few rounds of the the publish or peril game), limited freedom to do what you want (in large part due to games around funding and publish or peril).
The private sector now offers top talent what they used to go to universities for in terms of access to the best resources, freedom of exploration, a great talent pool of co-workers and decent working environment. Therefore is it any surprise people leave academia?
I and a few other posters mentioned this elsewhere, but at many universities athletics programs make profits that are able to be used elsewhere in the university. This often doesn't include donations on the order of a hundred million dollars at many schools for upgrades, etc.
As I read it, a vast majority of university athletic program are unprofitable. Many of them force their students to pay a special fee to fund their athletic programs. The following is probably not the best article but a start...
http://www.washingtonpost.com/sf/sports/wp/2015/11/23/runnin...
and being successful at one time does not necessarily guarantee later success. Mozart died alone and with no money. Turing killed himself at the age of 41. It's not hard to look for more recent examples.
Seems so; become a grunt programmer/admin and hope you don't get automated away too quickly (with my current academic performance I count myself to that group of people).
As for the second point: Depends on how you define success in life. If life's a paycheck maximization game to you, many science superstars weren't very successful...
What's wrong with this? Right now universities produce far more Phds than tenure track positions. If the free market is providing jobs for many of them, all the better.
The head of the CS dept of my undergrad once said to a corporate recruiter, "You can hire our undergrads, but we exist to turn them into Phds"
I started out college wanting to be a professor. That was fleeting, once I saw what my most respected professors and advisors were up against, grinding out papers, looking exhausted and subtly hinting that I should pursue anything but my initial goal.
When I started college, for some reason, I also thought the most interesting work was being done in academia, without exception. Now, a few years out, it is clear that the most interesting work is done at a crossroads - between top CS and engineering universities and industry. I do think it is getting harder to resist the lure of industry (based on the experience of my friends with PhDs), mainly because of comparable pay and far fewer peripheral responsibilities like managing labs, advising students, etc.
Modern AI is closer to math than the ad-hoc but often creative solutions of AI in the 70s/80s. This is a very good sign, but it brings AI closer to the kind of work that engineers do with MATLAB. I personally prefer it that way, but it's definitely quite different to other types of programming.
