Wouldn't this concern of yours also apply to the press under any definition? I can't afford to distribute my little pamphlet as much as the Times does their newspaper. And the constitution specifically protects the press, so you need a constitutional amendment. Rather than trying to make an end run around the constitution with a distinction about "money" not being speech, despite the obvious goal of the law being to regulate political speech itself via targeted restrictions on money.
Voting blocks are just simplifications of reality. Following that line of thought too far leads to bad arguments. The full truth is that any individual voter has a negligible effect on the outcome of an election.
Looks like it's just a matrix decomposition in the paper. I'm guessing anyway. These attention papers are always a painful mix of mathematical, quasi-mathematical, and information retrieval jargon.
There is something in the github repo about higher-order decompositions. Don't find where the method for factoring is given.
These are good arguments about phd's not going into academia. But a postdoc at a university is an underpaid training position for the most part. Not a worthwhile sacrifice if you want to work in the private sector in my field. Maybe it's more important for industry jobs in biology or something.
Personally I think it's a warped system that takes advantage of cheap labor from developing countries. And it feeds itself. The more temporary research staff professors can hire, the less permanent research staff universities need.
That's mostly reasonable. But for some people at least, the PhD doesn't give them all the skills they are looking for, so they might do another post-doc (2-3 years at slightly better than PhD lifestyle), before jumping ship.
They should have stayed rare, but they’re a way for universities to get free money from the government as well as a temporary (but sadly permanent) fix for academia’s deep structural problems.
Just to raise the number from 0, I did exactly that. The post-doc was in a research institute, a department that focused on usability. It seemed worthwhile to learn how to really do user tests and usability evaluations in practice, in the best way possible academically (or so I thought) but with real enterprise projects. Contrasted a lot with the quite theoretical PhD.
We can always argue about whether it was worthwhile ofc. Moneywise, definitely not (though the pay wasn't all that bad compared to industry compensation in my countries). But I learned what I wanted to learn, and the issues the position had were speficic to the place, not the concept of the position, if that makes sense.
A postdoc can also act as temporary employment when one can't immediately find a position outside academia. They tend to be very low commitment as a result of the laughable compensation.
generally you are somewhat right, but it's complicated by country. for example it would be less surprising that a UK-style 3 year PhD might want a couple more years training after.
> But a postdoc at a university is an underpaid training position for the most part.
That's not really true or fair. You're talking about a person who's done likely 4 years of undergraduate and between 4 and 8 (median 5?) years of graduate studies and gained a Ph.D. This person is without a doubt "an expert" in their field. Does that person still need training? Sure. But who doesn't?
You're also talking about someone who is probably in their early to mid 30's. Eventually people want to start having a life. And you're expecting them to make pennies? I'm not joking. Here's a list of CS Post Doc stipends[0]. Go to Berkeley or Stanford and make $66k-$80k? I've gotten more money as a PhD Intern! You're going to get more than double that in industry before we talk about RSUs.
I think this is something we need to take quite seriously here. There aren't really good reasons for people to choose this path, especially considering how disruptive it is to your life (have fun moving again in a few years :). Maybe we can hope that those who go are just really dedicated to research and following their passions. So dedicated that they are willing to put off normal human things like starting a family, maintaining relationships[1], or even just building one's net worth (or paying off student debt). Some people want to become a professor[2] so will do that.
But the big thing I see is that there's nothing to attract the best people. (We can extend this to even just professorships[3]). So who are becoming the postdocs and professors? Who are becoming the people that educate the next wave of people? If you are at the top, you likely aren't going the professor route, you're likely going to industry where you'll likely get 2x-3x the pay AND more freedom in your research. Even if you're pretty mid you can get 2x-3x the pay in industry with an honestly less stressful life (but you may not be doing research or likely have more restriction of what research you can pursue). This, seems pretty disastrous. At best, unstable. There's not a lot of professor spots but I think we need to have some incentive system such that people at the top are encouraged to carry on their research for the public as well as educate future researchers. As a business you want to be greedy and capture top talent, but that's also because you should be thinking much shorter term. At social levels, we have to think generations.
TLDR:
So are we confident we are getting "cream of the crop" (even if shared with industry) in academic positions? Are we sure that's even true at highly prestigious institutions?
I think the answer is "no", we should not be confident in that outcome.
[1] Good chance you met your partner while doing a PhD and yay now you both need to find post docs but they are competitive and so good luck finding one in the same place? Hope you like long distance relationships or are willing to let one partner make a bigger sacrifice.
[2] Which weird thing that we train someone to become an expert researcher and then as a professor we give them such high and diverse workloads that they will often have little time for research and instead will more be managers. The research skill degrades with time and only keeping up at a high level is not sufficient.
[3] In CS a post doc is also often a way to elevate your status. Like you got your PhD from a lower tier university and so doing a Post Doc at a much higher tier can make you employable at a more prestigious university. Because prestige still matters a lot, especially since it is tightly coupled with things like equipment access.
[side note]: I think with the competitiveness that it is a bit odd we still have these strong notions of prestige. Just to put things in perspective, if every graduating PhD at the number 1 school, they likely graduate enough people that you could fill all available faculty positions available. Given how prestige matters, it should be clear how high prestige graduates permeate into lower prestige positions. More clear when you start to consider things like how many top universities are in not the greatest living locations. https://drafty.cs.brown.edu/csopenrankings/placement-rank.ht...https://jeffhuang.com/computer-science-open-data/
Training is how the low pay is justified. I certainly agree it's not a good deal, more of a supply and demand effect. Are you saying I'm wrong about the definition? It's pretty universal: https://ncses.nsf.gov/pubs/nsf21327
I do agree it probably gets used more as an easy way to get experience, the usual thing organizations crave in their new hires. But it isn't the only way. Top graduates can get temporary research faculty positions or even go directly to tenure track. Or in STEM fields, a few years in industry is seen as positive. Industry and Government research labs also have their own "postdoc" positions which can be pretty much normal pay.
It’s not universal. Nsf is a narrow view of academia. There are other countries that pay postdocs properly (ie comfortable own place with kids) with benefits and pension.
> Are you saying I'm wrong about the definition? It's pretty universal: https://ncses.nsf.gov/pubs/nsf21327
Yes, I would say it is inaccurate to say that a post doc position is about training. Unless you're referring to them training undergrads and grads.
I think there are 2 key things we need to note about your source
>> Which items or job requirements are most commonly used by institutions to define postdocs?
This is in the context of __hiring__. So they are looking at listings. I don't think just because you are saying that this job will provide training means the job is a training position. Or what we'd commonly associate with a training program, a junior, or anything like that.
>> Is intended to provide training in research?
The wording here is important. Personally, if you asked if a postdoc was __intended__ for training, I would agree. But if you asked if I thought you should expect to be trained during a postdoc I'd be more hesitant. Certainly that depends. If I hired you as an L4 at Google, and said there would be training, what would that mean to you? Do you think it would be fair to say? What about at an L5 (senior)? L6? L7? L8? Certainly not at L9 or L10, right?
There's lots of things that have an __intended__ outcome but don't have the outcome. Momentum is a bitch and when we're talking about institutions, well momentum is a really powerful force. So I wouldn't take that source as a cout de eta for concluding that a post doc is about training the post doc. Honestly, I feel like a postdoc is closer to taking a L5 and saying we're training you to become an L6 or eventually L7. And certainly that's not a "training position" in the context of paying someone as if they are inexperienced. You can always gain more experience.
> Government research labs also have their own "postdoc" positions which can be pretty much normal pay.
FWIW, if you work at a government lab like LBNL or LLNL and take a __staff__ position then your compensation is going to be much closer to the base pay of a big tech. It's quite common to see these people jump ship and triple their salaries (sometimes they return. Often because it is nice to work at labs). I've seen people do this fresh out of PhDs, no postdoc involved. The postdoc positions pay lower.
It isn't just job ads. For half the schools there was a formal definition. Also it has this part: "In 2001, NSF and NIH established a formal postdoc definition as “an individual who has received a doctoral degree (or equivalent) and is engaged in a temporary and defined period of mentored advanced training to enhance the professional skills and research independence needed to pursue his or her chosen career path.”
And yes you should expect to receive training. In the sense that you have a right to it and the supervisor has an obligation to provide it. You seem to be mistaking my criticisms of these jobs for defenses of them.
> “an individual who has received a doctoral degree (or equivalent) and is engaged in a temporary and defined period of performing independent research.”
Would be just as accurate.
Btw
> “an individual who has received an undergraduate degree (or equivalent) and is engaged in a temporary and defined period of mentored advanced training to enhance the professional skills and research independence needed to pursue his or her chosen career path.”
Would be an accurate description of a PhD student too.
The official definition could also be accurate if describing an Assistant Professor (professor before tenure is granted)
So this is why I'm pushing back on the definition. Calling it a "training position" is just an excuse to have cheap labor and be exploitative.
Let me give an anecdote:
I was a PhD 1995-2001 at UCSF. I made $25K/year plus UC benefits.
Then I was a postdoc at Berkeley 2001-2004. I made $76K/year plus UC benefits plus the service clock for the pension started.
Then I was a staff scientist at LBL 2004-2007. I made $100K/year plus UC benefits and the service clock for the pension continued (I have enough service credit to get some thousands of dollars a month after I retire). It was an exhausting job and I did not enjoy having to simulateously publish, travel, and get grants to hire people to do research for me. I concluded this was not enough to buy a house and raise a family, so I left for industry.
I worked at Genentech 2007-2008 and made $120K ("Senior Architect") plus full benefits (which were pretty good).
I then moved to Google as an L5 (I had wanted to work there for ~10 years before I got hired) and started around $140K plus stock options, benefits, and retirement. Over the decade I worked there (2008-2019) my pay was increased signficantly every year, ending around $250K (including the year that Eric Schmidt gave everybody a 10% raise. thanks eric) along with ever-increasing options and then RSUs, worth millions (about $200K/year), excellent benefits, and saving a ton for retirement. There were a lot of side benefits- on the job training, free phones, etc. I really lucked out getting hired and promoted and being there during a growth period; I don't ever expect to have a role like that again.
This was the first time I truly felt like I could have kids and a home in the Bay Area without going deeply in debt.
The first few years at Google were stressful, I worked about as hard as I did at LBL, but every bit of hard work was compensated in some way or another- additional pay, access to resources (I basically could use all of the idle cycles in prod at Google- 1-3M cores- to do protein design, and publish). Every bit of research work that led to publications had far more impact (mainly due to the employer's position in research) than before, and I felt supported by the infrastructure to do ambitious things.
I spent a year at a Startup, making $250K and similar benefits to Google. I was employee #11 and have (worthless) equity- if they IPO I might get somewhere around $1M depending on how diluted my shares are and how big the IPO is. I am not allowed to sell the shares on secondary market.
Now I am back at Genentech and make more than I ever have in terms of base pay, while the stock (Roche stock) isn't nearly as valuable as what I got from Google. Fortunately, all those previous years helped build up a big buffer that will help pay for my kids to go to college (hopefully, some of my payment will go to help somebody else's kid get financial aid).
I'd actually love to return to LBL because if I do, my career will actually form a correctly nested path: LBL -> Genentech -> Google -> Startup -> Google -> Genentech (I am here) -> LBL. I'm sure I could get rehired there but at this point, why would I want to work harder for less money and publication credibility?
Looking back at my postdocs, while I learned a lot, I did not like the power dynamic with my PI, and really only persisted for 3 years with the goal of getting a much higher paying job.
I don't think the person you are replying to was expecting this. Rather, I read their comment as agreeing with you that the benefits and compensation were low enough that few people would be interested in the position except as a stepping stone into a better faculty position.
>You're also talking about someone who is probably in their early to mid 30's. Eventually people want to start having a life. And you're expecting them to make pennies? I'm not joking. Here's a list of CS Post Doc stipends[0]. Go to Berkeley or Stanford and make $66k-$80k? I've gotten more money as a PhD Intern! You're going to get more than double that in industry before we talk about RSUs.
Now maybe I'm out of touch, being a postdoc and all, but hasn't the tech industry been suffering mass layoffs this past year or so that really make it a bit unrealistic to label salaries above the American median as "pennies"?
This is the opposite of my experience. Big tech is far more hours and stress. And the work is much narrower and boring. The system is designed to make people into commodities and make you do work in the way that most benefits managers senior to yourself. Solve this narrow piece of a problem using this specific method alongside 5 other people doing other narrow pieces, versus a startup where you can just own the whole thing and iterate it faster.
Startups pay less and have to take what the can get. This probably causes a vicious cycle that makes the pay gap worse and worse.
IMO the stress is different. In startups you own things a lot more personally, and depending on how you handle stress that can be more challenging than the stress at big tech.
Hours can very much depend on the big tech and startup individually, but big tech is certainly moving in the wrong direction here, and in the typically big corporation/enterprisey worst way possible (increased employee monitoring, shallow and manipulable KPI's that act more as perverse incentives than actually improve anything and that increasingly show the disconnect between upper management and engineers)
> And the work is much narrower and boring. The system is designed to make people into commodities and make you do work in the way that most benefits managers senior to yourself. Solve this narrow piece of a problem using this specific method alongside 5 other people doing other narrow pieces, versus a startup where you can just own the whole thing and iterate it faster.
I agree and its my primary complaint/dislike for my job. I would love to, and probably will change to a startup or other job with a company I actually care about and a role I find valuable, but I expect the pay to be a significant drop when I do. Until then I'm leaning on the salary to reduce the amount of time I need to work overall significantly and putting in my time so to speak.
For now big tech is winning, but if startups and employee equity start having the potential payoffs they used to then it changes the balance of the above equation by quite a bit.
* If ByteDance divests their US TikTok operations, they create a new competitor that could potentially out-compete them in other (non-US, non-Chinese) markets.
* Whatever amount of money they get for this divestiture would be much lower than what the business is worth to ByteDance (when your options are sell or shut down, potential buyers will not feel the need to bid high).
* ByteDance's US TikTok operations are certainly of non-financial value to the Chinese government. That value is likely orders of magnitude higher than their financial value to ByteDance. Selling that user base is probably not preferable to shutting down. Influence campaigns are certainly easier to run on a platform you own, but certainly those campaigns are already running on Facebook, Instagram, Twitter, etc. Why add another platform that they can't control where they have to run influence campaigns?
They ban American companies from distributing or maintaining the application, not specific content. More of a "criminal conspiracy" type of behavior rather than something that would relate to wikileaks or whoever. Can as well argue what is being banned is the data-collection side of the operation since it's the part that depends on using the app. Those companies still have the freedom to publicly state whatever the CCP has to say.
The law indeed needed to be carefully written to "skirt" any first amendment violations, and SCOTUS unanimously agreed it had done so successfully.
"Content" is narrower than "speech". In that case the restriction was described as "content-neutral". Hence it didn't require strict scrutiny, only intermediate scrutiny. Which seems like the blueprint for how they wrote the TikTok law.
Of course the overwhelming majority of American do have one of those forms of coverage. You might as well argue American don't have housing or food either since most people need aren't on welfare programs to pay for them.
Over 90 percent of people on ACA plans get subsidies too. Also emergency treatment is guaranteed.
It's certainly a mess of a system, but every time the government does something to "fit" it, the price goes up faster and it becomes a bigger mess.
reply