In general our lab was very good at research & the rigidity of current academic institutions means that
1. admins are always the first hiring priority
2. academics do a ton of free work in publishing, service to their institutions on panels and committees, etc.
The leadership above our program head were so entrenched by block grants, & their de-facto recipients, that there would be fights over post-doctoral salary increases on the $5k level.
We had more money, which we won on competitive contracts, meaning if we won new contracts we could pay for better people... this was not welcome news to block grant recipients who had, essentially, fixed budgets & old mindsets.
The result is, as it typically is, driven by incentive structures.
There are a few different things defining the incentive structure:
1. Management by private institutions : National labs within the US are no longer managed by the government, there are LLCs which charge statutorily limited management fees of ~7% of lab revenue. The higher the revenue, or higher the operating budget, the higher the $ accumulated by the management company.
2. Liability limitation : Large science orgs, in the US, are deathly afraid of lawsuits related to personnel & the business of conducting science (ie employing lots of people who may test the boundaries of technologies, materials, & etc which could be hazardous long term). Typically government institutions are self-insured institutions, meaning there is no insurer who will underwrite their liability in the event of an incident. Safety & Admins are deployed as inertial dampening on the approvals of new research efforts. Admins also dilute responsibility for decisions made high within the orgs, via a hierarchy of admins that looks more like links in a chain: one admin reports directly to another admin, effectuating potentially arbitrary delays and dilution of scientific input on decision making within day-to-day operation.
3. Incumbency of older guard : Block funded scientists (the majority of whom were hired 25years or more ago) will get staff positions within large science orgs, because there's no risk to continued employment of that person long term to the org. Staff have all the political power at my old org, therefore staff positions [=] influence to drive the org. Older scientists are reluctant to cede political power, but also do not have dynamic budgets. Their task is to run projects with cutting edge science on, mostly, post-docs & grad students. Paying more for labor means they will have fewer papers, the scientific currency.
The incentive structures of those funding the university's research department. In many cases, state governments, who themselves are run by congresspeople responding to the incentive structures set out by voters. Those voters are themselves tactically voting to get the best possible outcome in response to the incentive structures set out by voting systems. And the people who decide what the voting system is are the congresspeople they're voting in, who have an incentive to keep the system as it is so they can stay in power.
In other words, it's incentives all the way down. At the national level, there are other factors that dilute the influence of this infinite regression of incentives, such as geopolitics, war, etc. But for states the only real factor that shines through - the green tint in our infinity mirror - is "how much money do we want to spend on secondary education" and the answer is invariably "less than last year" and "can we find someone else to pay for it?"
A related read is by Fei-Fei Li, who wrote more details about her point of view in a November 2023 article in The Atlantic [1]. The submitted article in The Washington Post focuses largely on the funding disparities, which she does highlight in her own article.
But her article in The Atlantic places a greater emphasis on the effects of a brain drain of AI researchers from academia to industry. She gives examples about how 40 roboticists left Carnegie Mellon for Uber in 2015, and how her close collaborator Andrej Karpathy chose OpenAI in its earlier days over "a faculty offer from Princeton."
She most notably recalls a quote said by someone at an OpenAI meeting that said that "Everyone doing research in AI should seriously question their role in academia going forward," and added the reflection: "To be honest, I wasn’t sure I even disagreed."
Uber left the self-driving car business four years later, selling that division to a startup. Did this work out well for the researchers, or did they end up getting a raw deal?
This trend heavily biases AI research towards Google problems. Perhaps things will swing away as LLMs (and especially smaller LLMs) take over the field.
I would not use publication numbers to argue anything. I'm an ML researcher at a university and publish at places like NeurIPS regularly.
There are so many edge cases that render this data meaningless. There are people who have their name on a dozen papers. Reviewers are terrible and papers with a lot of graphs run at large scale and easy messages tend to get in (I'm not saying that this is the only thing Google publishes, but people with access to a lot of compute tend to publish papers like this).
The funding gap between Google and academia is also massive. The fact that we're still competitive with Google is simply a testament to how incredibly efficient universities are and how inefficient Google is. They pay their people several times what we get paid. They have engineers which we don't. They have orders of magnitude more compute. They don't waste time on grants. Their total AI research budget is larger than the total combined AI research budget of all US universities. But.. somehow, they still account for only a small fraction of papers.
In the past when I was in academia, I saw a really interesting model: you have a faculty position at a university or a national lab. But you also set up an "Institute" which is independent from the U or lab. Instead it's an LLC owned by the faculty. The institute applies for grants ("wearing your institute hat") and the resulting grants are not subject to the university or lab overhead. Instead, that overhead goes to fund the admin structure for the institute, which can be far more minimal than a typical research institute.
With an approach like this you can get $2-5M/year in funding, hire 1-2 really good employees, and use the rest of the money to pay for cloud infra (or on-prem or other infra if that's your preference). And of course, any ideas you have, you spin off into startups that you have equity in.
Yeah, this is also good for donors. When you donate money to a professor's work through a university, the university ends up taking maybe a 30% cut. But if you donate to some independent institute that's controlled by the same professor, the university gets no cut.
It's annoying organizationally, though, since employees can't just hop between the sides. So it makes it harder to have a sane reporting structure, and encourages the pseudo-flat "nobody technically has a boss but also nobody is empowered to promote you" academic style.
So many insightful comments, I don't know this adds much but "basic cost of living" or "living reasonably" has gone up fairly exponentially in the last 10-15 years. Most people in academia, if they were on the fence at all about going to industry, are probably jumping.
In some not so distant past a family could live on a single-income, one person is an associate professor at a university type arrangement. These days without dual-incomes it's anomalous if you can even afford to buy a home.
It's a circular problem too because the recent waves of "tech boom" + after-shocks from CV19 and associated economic "policy" the slope has become even steeper.
It doesn't help that modern academia, if it was really ever different, at the graduate, post-graduate and doctoral level is basically indentured servitude. There's a reason why most graduate programs are filled with "foreigners", it's no different than H1B abuse in the tech industry -- unversities know they can abuse this work force for peanuts.
So if you just finished your PhD after doing 70-80 hr weeks for the past 4-6 years and you have choice A) Continue the same and hope you can get a crack at your own research team B) Get a job, work ~10hrs/day and afford to buy new clothes, a decent car, eat well and not live in a shack, you'd likely pick choice 'B'.
Has the cost gone up, or have expectations? None of my friends paid for subscriotions for anything but a landline, cooked moat of their food, and wore years old clothes a couple decades ago. People consume at a much faster rate today.
This might be a US-centric experience so I'm prefacing...
Basic cost of living is : rent, food, utilities. Those three have gone up astronomically, especially in the past 3-years.
A graduate stipend is probably $22K/year, maybe more for a PhD student. 20-years ago that was reasonable, if rent is $1700/mo you're already underwater given you have to pay taxes on that meager $22K.
In my graduate experience, I didn't see most grad students walking around with the latest cell phones, laptops or luxury goods, they struggled with the basics.
A number of graduate students at various universities are going on strike and forming unions and honestly more power to them:
Some of them are great but then I remember when I was floating around research labs some grad students would cry and tell you not to do what they were doing because its miserable. But then life has many fun horrors does it not?
It seems like SWEs are a dime a dozen now, and they will “come to you” - but I’m seeing more and more internships and roles seeking PhD grads/undergrads for roles that traditionally you may have seen just this as a nice to have or a plus.
I’m convinced these secret founders Signal group chats are all just CEOs saying how scientists and academia persons are really who they want working for them. My hunch is that they will most likely be less enthralled by the usual tech bubble that these companies tend to hire in, less problematic, and “less company focused” (focus on the task at hand, don’t lean into company culture or ask questions, follow the straight line).
This story is at least a few decades old. The only reason to stay in acedemia as a computer scientists/software engineer is if you want to focus on cutting-edge research or otherwise have a passion to teach others. Industry will outbid for everyone else.
Readit News
In general our lab was very good at research & the rigidity of current academic institutions means that
1. admins are always the first hiring priority
2. academics do a ton of free work in publishing, service to their institutions on panels and committees, etc.
The leadership above our program head were so entrenched by block grants, & their de-facto recipients, that there would be fights over post-doctoral salary increases on the $5k level.
We had more money, which we won on competitive contracts, meaning if we won new contracts we could pay for better people... this was not welcome news to block grant recipients who had, essentially, fixed budgets & old mindsets.
The result is, as it typically is, driven by incentive structures.
1. Management by private institutions : National labs within the US are no longer managed by the government, there are LLCs which charge statutorily limited management fees of ~7% of lab revenue. The higher the revenue, or higher the operating budget, the higher the $ accumulated by the management company.
2. Liability limitation : Large science orgs, in the US, are deathly afraid of lawsuits related to personnel & the business of conducting science (ie employing lots of people who may test the boundaries of technologies, materials, & etc which could be hazardous long term). Typically government institutions are self-insured institutions, meaning there is no insurer who will underwrite their liability in the event of an incident. Safety & Admins are deployed as inertial dampening on the approvals of new research efforts. Admins also dilute responsibility for decisions made high within the orgs, via a hierarchy of admins that looks more like links in a chain: one admin reports directly to another admin, effectuating potentially arbitrary delays and dilution of scientific input on decision making within day-to-day operation.
3. Incumbency of older guard : Block funded scientists (the majority of whom were hired 25years or more ago) will get staff positions within large science orgs, because there's no risk to continued employment of that person long term to the org. Staff have all the political power at my old org, therefore staff positions [=] influence to drive the org. Older scientists are reluctant to cede political power, but also do not have dynamic budgets. Their task is to run projects with cutting edge science on, mostly, post-docs & grad students. Paying more for labor means they will have fewer papers, the scientific currency.
In other words, it's incentives all the way down. At the national level, there are other factors that dilute the influence of this infinite regression of incentives, such as geopolitics, war, etc. But for states the only real factor that shines through - the green tint in our infinity mirror - is "how much money do we want to spend on secondary education" and the answer is invariably "less than last year" and "can we find someone else to pay for it?"
But her article in The Atlantic places a greater emphasis on the effects of a brain drain of AI researchers from academia to industry. She gives examples about how 40 roboticists left Carnegie Mellon for Uber in 2015, and how her close collaborator Andrej Karpathy chose OpenAI in its earlier days over "a faculty offer from Princeton."
She most notably recalls a quote said by someone at an OpenAI meeting that said that "Everyone doing research in AI should seriously question their role in academia going forward," and added the reflection: "To be honest, I wasn’t sure I even disagreed."
[1] https://www.theatlantic.com/technology/archive/2023/11/ai-et...
There was also past discussion on HN about her article at: https://news.ycombinator.com/item?id=38362242
https://news.ycombinator.com/item?id=9638121
https://news.ycombinator.com/item?id=9602655
This trend heavily biases AI research towards Google problems. Perhaps things will swing away as LLMs (and especially smaller LLMs) take over the field.
There are so many edge cases that render this data meaningless. There are people who have their name on a dozen papers. Reviewers are terrible and papers with a lot of graphs run at large scale and easy messages tend to get in (I'm not saying that this is the only thing Google publishes, but people with access to a lot of compute tend to publish papers like this).
The funding gap between Google and academia is also massive. The fact that we're still competitive with Google is simply a testament to how incredibly efficient universities are and how inefficient Google is. They pay their people several times what we get paid. They have engineers which we don't. They have orders of magnitude more compute. They don't waste time on grants. Their total AI research budget is larger than the total combined AI research budget of all US universities. But.. somehow, they still account for only a small fraction of papers.
With an approach like this you can get $2-5M/year in funding, hire 1-2 really good employees, and use the rest of the money to pay for cloud infra (or on-prem or other infra if that's your preference). And of course, any ideas you have, you spin off into startups that you have equity in.
It's annoying organizationally, though, since employees can't just hop between the sides. So it makes it harder to have a sane reporting structure, and encourages the pseudo-flat "nobody technically has a boss but also nobody is empowered to promote you" academic style.
Deleted Comment
In some not so distant past a family could live on a single-income, one person is an associate professor at a university type arrangement. These days without dual-incomes it's anomalous if you can even afford to buy a home.
It's a circular problem too because the recent waves of "tech boom" + after-shocks from CV19 and associated economic "policy" the slope has become even steeper.
It doesn't help that modern academia, if it was really ever different, at the graduate, post-graduate and doctoral level is basically indentured servitude. There's a reason why most graduate programs are filled with "foreigners", it's no different than H1B abuse in the tech industry -- unversities know they can abuse this work force for peanuts.
So if you just finished your PhD after doing 70-80 hr weeks for the past 4-6 years and you have choice A) Continue the same and hope you can get a crack at your own research team B) Get a job, work ~10hrs/day and afford to buy new clothes, a decent car, eat well and not live in a shack, you'd likely pick choice 'B'.
Basic cost of living is : rent, food, utilities. Those three have gone up astronomically, especially in the past 3-years.
A graduate stipend is probably $22K/year, maybe more for a PhD student. 20-years ago that was reasonable, if rent is $1700/mo you're already underwater given you have to pay taxes on that meager $22K.
In my graduate experience, I didn't see most grad students walking around with the latest cell phones, laptops or luxury goods, they struggled with the basics.
A number of graduate students at various universities are going on strike and forming unions and honestly more power to them:
https://www.latimes.com/california/story/2023-01-02/uc-strik...
It seems like SWEs are a dime a dozen now, and they will “come to you” - but I’m seeing more and more internships and roles seeking PhD grads/undergrads for roles that traditionally you may have seen just this as a nice to have or a plus.
I’m convinced these secret founders Signal group chats are all just CEOs saying how scientists and academia persons are really who they want working for them. My hunch is that they will most likely be less enthralled by the usual tech bubble that these companies tend to hire in, less problematic, and “less company focused” (focus on the task at hand, don’t lean into company culture or ask questions, follow the straight line).
Dead Comment