When the A.I. Professor Leaves, Students Suffer, Study Says | Hacker News Reader

When the A.I. Professor Leaves, Students Suffer, Study Says

nytimes.com

91 pointsOsiris30103 comments

Pay the professor more money. Top Universities are sitting on tens of billions of dollars of endowments, and are willing to pay football and basketball coaches millions of dollars. They are not insulated from market pricing.

In many cases, a new assistant professor's salary at a top university is lower than than entry level salaries for software engineers in technology. For considerably higher stress, harder work conditions, and less job security until they make tenure.

Compare this with medical doctors at teaching med schools: they are often paid slightly less or competitive with practicing doctors.

I am a research scientist with a PhD, and while academia has perks, I'm very happy I went to industry.

Yes, this exactly. I'm an assistant professor who's fine with my salary and position, but I can see firsthand how difficult it is to recruit and retain excellent colleagues.

But why would the university bother?

The students are paying to go to <top name school>, not (very often) because they want to be taught by <top name professor>. From the university's perspective, there's little financial incentive to retain that top talent. Making it worse, the professor isn't being stolen by a competitor (another university) where students might spend their money instead.

In short, if you're asking for the free market to solve this problem, it has- from the professors' perspective.

Edit: put another way, these professors can earn the private industry far more money than they can the university. Often a lot more. Salary competition between such pairs of actors will always result in the same thing.

If universities are profit maximizing entities why do they get preferential tax treatment?

Paying employees more is not profit-centric... I'm not sure about your argument here ..

University athletics is certainly a profit center, but research and general operations is devoid of cash, even at R1 institutions. I struggle to find money for salary and most private industries will offer 20% more for any employee I hire

Quoting the grandparent post:

If that’s how a university makes decisions then I don’t think it ought to get tax advantages.

The top schools aren't the ones who need to worry, and it's not the famous professors who need to be retained.

It's the mid and low ranked schools who are facing competitive pressures from alternative options.

The trouble is that the market rate for professors in other subjects is not as high, and they get mad if the differential is too great.

Is there evidence for this?

You should look up salaries for professors/lecturers that teach business, law, and medicine. They pay a lot more than salaries for professors/lecturers for history, math, and literature.

Med is different for a lot of reasons.

But, yeah, the delta between econ/law and CS faculty salaries is why I turned down TT offers and went to industry.

They can pay. They're choosing not to.

Just a consideration: Do these differences in salaries for professors of the different subjects also show up in tuition fees for students of the respective subject?

Yes. https://www.pewtrusts.org/en/research-and-analysis/blogs/sta...

Not at the types of institutions I would be working at (or most institutions, for that matter).

They have their own schools which is why they can do this. These are considered graduate programs as well.

So does CS at a lot of places.

Also, econ/fin/law makes more even at undergrad only institutions. It's a perception thing with admin. ESPECIALLY at those places, until institutions pay CS profs market wages, stay away unless you want to wake up when you're 60 and realize you can't afford to retire like all your grad school friends and so can't spend as much time as you'd like with your grandkids because you're teaching courses every semester.

Consider the plight of the medical student though. Worse than that of a CS grad student, that's for sure. Made to practice medicine on people while sleep deprived as a form of industry-approved hazing, a practice that doubtlessly has a substantial death toll associated with it.

Never understood why this doesnt get any press. Its absurd

Gets lots of press. Google doctor interns lack of sleep for a litany of deeply-researched news articles.

Hospitals are being privatised and private universities dominate medical education. Where’s the political pressure point?

It gets press but only so many ways to report the same story for years on end.

It gets press, but it also a legitimate training technique in many fields. There is something to be said for living and breathing nothing but your job for a period of time. Stress and the ability to perform despite stress is also something that must be learned through experience. Doctors, attorneys, the entire military ... the juniors are run through a few years of long hours and horrible conditions. The downsides have to be weighed against the training value.

The military is trying to mold you the same to make you replaceable and obeisant. Doctors don't need to be all the same.

Replaceable and obeisant? You've watched too many movies. Skills learned while under stress will be with you when you are under stress again later. A doctor is trained under stress so that in the future, when their patients are about to die in front of them, that they react as trained. A doctor will see more blood, more death, than most soldiers. The parallels in stressed training are not insignificant, and have nothing to do with obedience.

That's just not true. Being trained in a relentlessly stressful environment means that you're less likely to be creative, thoughtful, and empathetic in the solutions you come up with. It can also lead to learned helplessness -- a sense that things must be done the way they've always been done, because there's just not enough time and space from others to do things properly.

During residency they are dealing with patients dying in front of them. The stress is already there. I don't see why being sleep deprived is a benefit.

Basic training is literally about making them think and act certain way. It is literally about them all teaching exactly same set of super basic skills so that the soldiers are replaceable by the end by each other. There is reason why soldiers personality changes after basic training (that is what their families and partners say).

More complex skills are not taught in basic training, but later in specializations.

Under stress and duress and lack of sleep, people learn less. Moreover, while soldiers do in fact have to work in difficult physical conditions in action, doctors don't.

And I am not even sure what death on itself has to do with it. Conditions of encountering it are incredibly different in both professions.

“...Moreover, while soldiers do in fact have to work in difficult physical conditions in action, doctors don't. And I am not even sure what death on itself has to do with it. Conditions of encountering it are incredibly different in both professions.”

Try working in an inner city ER at night when it can often feel very similar to a war zone. Instead of sandbags and soldiers with rifles we’re surrounded by bulletproof glass and police officers.

That is nonsense argument. Doctors don't sleep in barracks or on grass, they go home to families. They eat what they choose to eat. They don't dig latrines. They don't have periods (days) of boredom full of made up work combining with short intensive assignments. They can leave the job.

Soldiers are not surrounded by bullet proof glass nor by police officers.

You still need sleep to form long term memories and make good decisions. Sleep deprivation is inhumane to the students and deadly to patients. There is no excuse.

You do realize that this isn't the entire training scheme? This is one stage of a multi-year process, starting long before med school. The most effective thing for patients would be to never be treated by trainee doctors but that isn't a practical option.

Yes. What's your point? Hopefully it's not that since patients are going to be treated by students, we may as well give up and have them treated by sleep-deprived students, or that sanity in one phase of the training process excuses insanity in another. But if you're not trying to make one of those arguments, I don't know how your comment is relevant.

> Consider the plight of the medical student though.

While not as bad, ML/NLP/CV/AI grad school is still pretty bad. Maybe 60-70% of what med students (more specifically, residents) go through.

Graduate school is probably worse because it's not time boxed and what you're trying to do may not actually be possible. When you leave your medical residency for the day you don't have the option to continue working.

Wife is in surgery residency, they do a lot of research projects, there is a lot of work to take home

It never ends. Wife is an attending surgeon, and she's currently working on refining a grant proposal at 9pm on Sunday =).

>While not as bad, ML/NLP/CV/AI grad school is still pretty bad.

Uhhhhh what? I'm in grad school right now. I feel... a lot better-off than any medical student or resident I've ever met, or even than my wife, who's a high-school science teacher (God help her!).

No way. It's not even close. And I say this as a PhD candidate in one of these fields at Stanford. Grad school is what you make of it, to a large extent. There's definitely toxicity, but it's not as systematic or ritualized to the extent that it is in medicine. In grad school, you have wide leeway over who you work with, and you can distance yourself from the toxic personalities if that's what you want to do without taking a career hit. That's not really possible in medicine.

Yeah, med school is the only “brainy” career path that endures military-style hazing on a ritualized and widespread basis.

To be precise: med school is (comparatively) a walk in the park. Years 3 and 4 of med school are a sort of "residency lite" where you rotate in different specialties, but the doctors don't really expect you to be useful, so it's fairly low-pressure other than odd hours. Study up so you can know what you're talking about, don't get in people's way, and you'll be fine. Residency is when the 80+hrs/week institutional hazing happens.

Yes, “medicine” would have been the correct term.

At a top tier university, sure. Try a department with 20% faculty turnover where all the junior faculty leave before the third year review as soon as they get their NSF CAREER award and you don’t know if you’re getting paid in the summers.

Currently getting a PhD at respectable grad school, my Girlfriend is currently at medical school. She has it way harder than I ever have, medical school is just on another level.

Excessive working hours are used as an excuse by medical cartel to justify practices that drive down the supply of providers through licensing requirements, and drive up salaries for doctors and everybody else up the chain. "See how hard we are working and how long we had to stay in school? We deserve to charge $500 for a 10 minute visit". "An acne cream that Europeans get over the counter for $5? We studied for 10 years and we know best what's safest for patients, so they HAVE to go to us first to get a $300 prescription." This is literally the arguments that medical associations have been using to lobby state and federal agencies to keep the status quo.

What needs to be done is a dramatic shift of power away from doctors in the medical industry. Flooding of medical labor markets with mid-levels - nurses, PAs, RAs, PTs - and relaxing respective scope of practice laws. Physicians will fight to death against new competition, even if it means killing themselves with unreasonable working hours.

https://www.auntminnie.com/Forum/tm.aspx?m=581583

"Don’t teach your mid levels procedures... Ultimately, a mid level that can do procedures, even easy ones, are your competitors."

https://www.politico.com/agenda/story/2017/10/25/doctors-sal...

Excellent comment. The AAMC and AACOM quite literally fit the definitions of cartels, artificially restricting the supply of medical doctors.

Also, it's the 21st century - why do doctors (more specifically, hospitals) generally work 9-5 M-F? Viruses and bacteria don't take the weekends off. Not to mention when people need to see a specialist, they need to wait weeks or even months for an appointment. Why? It's absurd. With a larger supply of doctors and something like 7am-10pm daily hours, it would result in a massive net benefit to society.

"The AAMC strongly supports bipartisan GME legislation introduced in both the House of Representatives and the Senate, the Resident Physician Shortage Reduction Act of 2019 (H.R. 1763, S. 348)" - https://news.aamc.org/for-the-media/article/gme-funding-doct...

“...it's the 21st century - why do doctors (more specifically, hospitals) generally work 9-5 M-F”

What country do you live in? In the USA hospitals are open and staffed for emergencies 24/7/365. I’m a Radiology Technologist who works in an ER and Surgery- and I’ve never worked less than 48 hrs per week. I work 2300hrs-0700hrs. If you are talking about viruses and bacteria, that’s usually not an emergency issue and yes, you CAN wait until normal business hours (unless you’re septic)

I live in the USA. Oh, you mean the same ER that will slap you with a several thousand dollar invoice (only after the services are provided, of course) after a 3 hour wait? Well that's a related but different story I won't get into at this time.

If I have have an infected tear duct (in fact, I do) requiring a visit to an ENT specialist (for example), it's frankly absurd that the next appointment requires a 2-6 week wait. And then it also affects the work day because ENTs docs are only available 9-5.

Why aren't there two shifts of doctors: 7am - 3pm, and (as an example) 12pm-8pm? At least I work a computer-based job; what about a construction worker or other blue-collar worker who has to take an entire day off because they can't log in and continue work remotely (and if you know blue collar folks, this is the reason many will skip preventative care until they are critically ill)?

Your argument seems to be "well you won't lose the eye or go blind so it's fine" and that simply isn't how we should be viewing healthcare.

And just for posterity's sake, I've spent years at hospitals and medical device companies, my background is in biomedical engineering, and I'm not what you'd consider a layman in this area.

"Flooding of medical labor markets with mid-levels - nurses, PAs, RAs, PTs - and relaxing respective scope of practice laws"

Well, I don't have any insider knowledge, but from a healthcare consumer perspective, it sure seems like doctors have gotten scarce and NPs are doing everything these days.

I don't know how the money flows in a group practice with one doctor and everybody else is not, but I'd be curious about how it works.

It's done to keep supply down. Medical costs are high partially because of lack of doctors willing to put up with this. Talk to the AMA.

While we are at it, pay the grad student more as well. Otherwise, don't be surprised that your good students end up all being foreign-born, motivated by immigration benefits, and local ones would [often wisely] be totally disinterested in the grind.

It shocked me when Republicans wanted to tax grad students... Seriously, 18-25k take home pay is poverty level

Grad students are taxed.

I believe the debate was around taxing the "tuition waiver" component of grad students, which grad students never see and is basically moving money from one pocket to another at the university.

getting university professors into an arms race with the salaries of the private markets would entirely shift the character of public education.

If universities had to play by the cutthroat logic of the market as a consequence research or teaching that is not profit oriented would have to die out, or tuitions rise to levels that put immense pressure on individuals and the public purse.

This is a dystopian scenario where academic freedom, basic research, and the non-commercial nature of research is completely lost. This is not something we should be going for, instead we should appeal to highly trained researchers to maybe think about the public good that is at stake here. People like Bengio show that it is possible. A future where essential AI research is commodified is a nightmare scenario.

How is this a nightmare scenario? Why is teaching being profit oriented a bad thing?

There is no necessary loss of academic freedom. Paying a competitive salary doesn't make research commercial -- the university is trying to retain researchers not turn a profit on the research.

> the university is trying to retain researchers not turn a profit on the research.

Depends on the university. Most claim ownership of researchers' inventions and have ambitions of profiting from them (many stumble at the commercialisation/monetisation step). For example, Stanford has reportedly made hundreds of millions off Google's PageRank algorithm, because Stanford owned the patent, not Larry Page. That's a good incentive not to work as a university researcher.

I think Larry Page did okay in the end. It's not like Stanford takes hundreds of millions from every rinky dink idea thought up by it's students.

Another way to look at it is researchers are getting somewhat angelic investments from schools.

The money to pay people comes from loans, which are thousands of people hoping they make more money in the future to pay salaries today. If salaries were higher, tuition would be even more difficult and raise the loans even more.

Really universities need to find ways to capitalize on their knowledge base to help pay salaries. They also need to reduce overhead... Many schools have several disjoined IT and sometimes HR departments.

Or we could cut out a few University Vice Provosts of Telling Professors to be Game-Changers who make $400k/year.

this is a straw man, no better than the argument that the NCAA and other sports organizations use to excuse paying so-called "amateur players" absolutely nothing, for, and I quote, "considerably higher stress, harder work conditions, and less job security until they [go pro]". parent never argued that professors need to be paid as much as private, merely that the salaries need to be comparable rather than several times lower.

They're in that race already because private enterprise exists. Either they lose or they adapt. Losing is having advanced research and tech be siloed inside massive corporations. Adapting is out-competing the massive corporations for the best, then being able to capitalize on the research.

Can't agree more. I was initially interested in academia and went pretty far down the graduate track but ultimately felt like the academic world just hangs professors out to dry when it comes to compensation (tenure is a poor substitute).

To be fair, football and basketball coaches bring in more money to the school. Paying a teacher more doesn't necessarily equate to more money in the school's pocket.

Football and basketball coaches likely help the school lose money. Almost zero US college sports programs are profitable[1].

Also, since when does a university have to justify its teaching quality by its profitability? That's the reason the school exists and why it has non-profit status...

1. http://www.ncaa.org/about/resources/media-center/news/athlet...

*directly. Sports programs help with recruiting and alumni fundraising. Both of which are profit centers.

How do you study that exactly? How can you know that your losses on sports are covered by your gains in recruiting and fundraising?

All the wealthiest schools are known for their academics, not their sports.

Schools recruit and fundraise betterin years with/after more successful football seasons.

If you know how you fundraise at 6-6 vs 11-2, you can quantify the value of a coach who brings you 3 extra wins each season.

Cool, and they wouldn't be doing anything without the school, so the money should go to the school's general fund.

Wouldn't it be cheaper to hand out more tenure, or at least eliminate adjunct and ensure that everyone has at least health insurance? As a start?

Medical doctors at universities and medical schools generally make 50% of salaries for private practice.

All the money is used to line the pockets of ever growing ranks of 'Administrators and Executive assistant' who are busy creating silly rules and pointless systems ;)

Probably can't pay the professors more money -- they're delivering that much value. But they can woo the professors back with flexible teaching arrangements that allow them to continue to teach while working.

It's mostly not about money - though it's certainly a piece.

It's about actual capacity to make an impact.

At a major company as an "AI" practitioner, you have immense amounts of data to play with, a huge budget and (relatively) little oversight.

As a researcher, you have to teach classes to undergrads, write grant proposals and scrounge around for data and project funding.

Which do you think you'd rather do if you care about making functional progress in your field?

The blog post "You Cannot Serve Two Masters: The Harms of Dual Affiliation"[1] from prominent AI researchers Ben Recht, David A. Forsyth, and Alexei Efros also highlights the challenges of co-employment by big tech companies. Well worth a read.

[1] http://www.argmin.net/2018/08/09/co-employment/

Pretty much all of the highest paid state employees are medical professors.

https://salaries.texastribune.org/highest-paid/

Even normal engineering professors are make 200-300k + they can still do consulting. Being a professors at a good university helps you market yourself and gets you those high paid side gigs.

That’s pretty incredible - #100 of the top 100 best paid still makes $652,640.

However, CS seems to be more in the sub-$200k range with only a few over $200k. That’s pretty easy to exceed in the industry and why people are likely choosing to leave.

Medical professors usually make their money by also working as doctors providing medical care at university hospitals. The university takes a cut.

A combination of surgeons and football coaches. What a weird dataset.

I actually think a lot of Universities in America are hiring bottom-of-the-barrel programers for their tenured positions as a result of the income difference from University to Industry.

I went to a top-10 CS school as well, and saw it happen several times. Professor that is actually really talented, educated and capable of articulating advanced CS concepts ends up leaving to be a sr. engineer at G, FB, MS, etc. due to differences in pay.

We lost our best math prof, and our best CS prof in my 4 years because of this. They both got replaced with imports from other countries that couldn't teach as well due to language barrier but probably where happily willing to accept the visa sponsorship and shot at USA life (which I understand).

"Plundering" and "poaching" are inherently wrongheaded notions of gross entitlement from employers expecting unilateral loyalty.

The problem should be approached as "how can we attract and retain the best" not thinking of them as property and their competitors as thieves.

I can’t even get software developers to think I’m terms of teamwork (Mostly I can filter by ones that are sympathetic). Imagine trying to get academics to do so.

All this reminds me of the late 1990s, when 50% of academics who knew HTML ran to a dot-com. It didn't last, and a few years later an awful lot of former professors begged for their jobs back. Some of them got it, but most didn't.

I've noticed a trend of basically any PhD who knows Python getting hired to join any of the many growing "data science" teams in large companies. Despite having PhDs (usually in non-related fields like biology or even education) most of these people know very little about engineering, statistics, CS and surprisingly science. All they know how to do is stitch together a bunch of keras code and try to get some arbitrary cost function drop.

I immediately realized that we are definitely in some kind of bubble, and that last time I saw this was the dotcom era. The new HTML is python + some NN library. As soon as the next recession really hits companies are going to realize that the majority of these data science projects are just huge houses of cards that provide questionable value. I honestly miss the post-dotcom era, where nearly everyone in tech was genuinely interested in the field and solving interesting problems.

I guess enjoy the fact that salaries are high and work is stupidly easy to get now, I suspect the party will be over pretty soon.

> Tech companies disagree with the notion that they are plundering academia. A [typical large dotcom] spokesman, for example, said the company was an enthusiastic supporter of academic research.

On occasion, I've asked multiple professors why they don't criticize particular dotcoms or industry practices, or warn undergrads about the nasty cultures of particular companies, and they've said they need the funding from those dotcoms.

Personally, I'd love it if undergrads learned a sense of professional ethics and objective understanding of current industry before they applied for internships, and, consequently, anyone who went to one of the worst (but best-paying) ones for a summer would have greatly increased difficulty getting dates when they returned.

I (soon to be non-undergrad) got the impression around 2016 that the admiration was wearing off of our classmates who did co-op terms at FAANG. One of our peers had done a term at a Cambridge Analytica subsidiary in town, AggregateIQ, so when the news came out, it was especially prominent in (at least) my mind. I can't comment on whether any romantic efforts were impeded or not.

> Personally, I'd love it if undergrads learned a sense of professional ethics and objective understanding of current industry before they applied for internships

Even with an understanding of ethics, undergrads, new grads, professionals and startup founders _usually_ pick the "unethical" option.

- Facebook knows this (in 2019) and thus offers more money to offset it. Several of my grad school friends from MIT/Stanford/Berkeley chose to work at Facebook because "they gave $300K-$400K total yearly compensation."

- US-based start-ups routinely sell products, give up equity control, or sell directly to the Chinese government or investors. Many times precluding them from selling to US customers/government. My startup recently closed a $20M computer vision deal with a Chinese customer related to person re-identification. Unethical? Probably to someone, yes.

The reason for all of this? Because "everyone has a price." If you believe your ethics can't be compromised, well then you haven't been offered enough money.

You realize the Chinese government is going to use your tech in their concentration camps and you don't care at all?

> My startup recently closed a $20M computer vision deal with a Chinese customer related to person re-identification. Unethical? Probably to someone, yes.

So... helping the Chinese government surveil and repress its ethnic minorities?

Perhaps those academic institutions should simply offer tenure much faster if they want to keep these professors.

It is hard to compete on money with the industry, but this is something that universities can offer.

From the article:

At Carnegie Mellon, 17 professors, all of them tenured, have moved into industry

Not sure about other universities, but it doesn’t sound like tenure made much difference at CMU.

I think this is not new in CS-ish university departments. With the start of dotcoms, it seemed a lot of professors and departments wanted their students to do startups, and the professor/dept. would get a cut of it.

This affected time spent on students, affected what some wanted students to work on (use the uni as a startup incubator, rather than research), and affected how some selected students.

I also recall professors who were almost no-shows, spending most of their time on their students' startups, didn't actually teach their one class for the term, and even their PhD students complained about having trouble getting meetings.

I don't know that some CS departments ever recovered from that, and it seems some are permanently in pursuit of commercial spinoff riches mode.

What’s the study’s control group? If we compare the number of companies founded by students over different time periods it doesn’t say much about the effects of the “brain drain”.

First we should abolish the 5-7 year PhD program and make it three years. Second universities should use this crisis as an opportunity to rethink Their approach to entrepreneurism. Third this actually creates opportunities for students to become professors. In most fields good luck!

Reducing the length of PhD programs (in the U.S. -- PhDs in Europe tend to be shorter) would have the side-effect of having even more PhD grads (of lower quality) that can't find a job.

Getting rid of tenure would facilitate mobility and increase accountability.

So keep them out of the workforce longer?

PhD programmes are 3 years in Europe because the education beforehand is much more focused. You've pretty much chosen what you want to do at at 16 and from 18 years old onwards you only study one subject. To be admitted to a PhD programme usually requires either a masters degree or a lot of reserach experience. Then the actual PhD programmes themselves are just research, no classes or teaching.

This is a broader trend in academia, have seen it happening in life sciences and medicine as well.

It isn’t just about salary. It is more about working conditions and job security. A university can never match private industry on those things, even if it could match the pay.

Tenured professors have vastly better job security than anywhere in the private sector.

Is there a way to access nytimes without signing in?

Use archive.is, e.g.

> http://archive.is/2qktN

clear cookies whenever you see the paywall. works for most news websites in my experience

Disable JS.

Hit the stop button on your browser at the right time. After HTML loads and before JS. Really! It’s not hard to time it, just did it on my iPad.

I really wish Google wouldn’t surface paywalled links (coming from a print NYT subscriber for over a decade)

Reply on news.ycombinator.com