Live data from Hacker News

PhD Simulator

research.wmz.ninja

161–170 of 285 posts

Re: PhD Simulator

#161
post #154

Earlier quoted context omitted.

> Classwork is basically dead-end work and the more you focus on it the less you have to show for yourself when trying to convince an advisor to work with you. nearly half of my year didn't get this and had to master out when we got to quals.

You mean they passed quals but couldn't get an advisor? Or failed quals?

Are those different things? At departments where quals have high failure rates, it's really more of an annual layoff than anything else.

In many programs, the department aims to admit far more people than will pass the quals. They need the Calculus and Pre-calculus TAs but do not have the advising capacity.

Even if everyone gets a 95% on the quals, the majority will "fail" by necessity because the department simply does not have the advising capacity for the number of TAs they need. Of course, the department typically designs the quals to these needs either explicitly or implicitly.

This is usually at least implicitly understood by the faculty, who will navigate it when absolutely necessary. For example, I've seen it happen that if a professor really needs a student and vouches for/protects them (eg because the research is computational and the student came from 5 years at Google), then the student gets more goes at the plate on quals than is typical.

Re: PhD Simulator

#162
post #54

Earlier quoted context omitted.

One dynamic I experienced that also isn't in the simulation: if you focus too much on classwork early on in order to pass your RPE, it can actually be hard to find an advisor. Classwork is basically dead-end work and the more you focus on it the less you have to show for yourself when trying to convince an advisor to work with you. Your goal should be to optimize for doing just well enough to pass your classwork. Als…

This would be incredibly bad advice in half of Physics and most of Math. An adviser would simply not trust a graduate student with middling grades to be competent enough to work with.

IME it's more that the advisor doesn't trust the student to make it through the annual layoffs (quals culling), and only wants to invest in people who they know will be around long-term.

At least in the poor (and honestly mostly useless) parts of Mathematics. Maybe Physics is less poor.

(Fortunately I was in CS, where the research output is actually needed by society and usually not pure masturbation, so the attitude toward coursework was "do well at what you need, enjoy what you want, and ignore what you don't need or want"

Re: PhD Simulator

#163
I think the most realistic thing here is the luck element. You have people around you passing and making it look easy and saying to do exactly what they did and those people make it all out to be a skill game. But you follow their exact method and still fail. And keep failing. Making you think there's some secret sauce that they aren't telling you about. But in reality the difference was just luck. That in one game you can slack off half the time and graduate just fine and the other half of the time you can't even get a single paper submitted. The tyranny of the stochastic system is probably one of the most damning things in a PhD.

Re: PhD Simulator

#164
lol I wish it was this easy. I got through the simulator in 6yrs 11mos on the second try. At no point was hope above 40%, except once early on (ended at 33%).

The funny thing is that I had 1 conf paper, 1 major result, and 1 figure left over. That's a good year extra, so I assume a perfect game would be to get the 3x papers and GTFO (which is the second best outcome, after not enrolling). There were a couple folks I knew that made it out in 5 years, but more that took 7+. Our lab was notorious for taking over 10, which I skirted by.

Like others said, this was lacking outside events (social/political junk). Hopefully version 2 will take into account: at least 1 family death and 1 additional tragedy, at least two months lost to helping or waiting for help from another grad student or post doc (they did have the lab equipment breaking, which was good to see, but missed the lobbying for every little purchase), at least one scope change, a half dozen favors to gain some political cache, a few experiments and/or rewrites to satisfy faculty members that just read about a technical issue they should have known, but didn't so they're highly sensitive to it, at least 6 months of arranging the data/results in a way that faculty can understand, 3 months of arguing that the lab standard procedure for some basic component is a decade out of date, a few months worth of preparing premature data for unnecessary meetings, one (and it better be just one) instance of an offer to help getting waaaay out of control, the hope boost after your first big conference and subsequent conference hope drops, the drops with each thesis defense from folks a year younger, etc. There's more, but that's off the top of my head. Oh, and that slight boost in hope when you hear someone else has a worse problem than your current one. That's a fun one.

Tip for those interviewing - ignore all the year 1-3 folks. 1 and 2 are basically undergrads plus some extra classes. 3 probably hasn't hit the first pile of bullshit yet. Find a year 5 or 6 in your field and talk to them alone. There's a reason they generally don't have senior grad students at recruiting events, and it isn't because they're too busy. Talk to them long enough to get to their exhausted attempts to rationalize some aspect of the experience. If their demeanor doesn't change, you might be safe. If they start hemming and hawing, that's a problem. They haven't even gotten to a specific, non-personal problem and they're having trouble keeping up the facade. The layers are: 1) Hey, social event, I get to take my mind off lab problems. 2) Getting a little boost by talking to someone still excited. 3) The quiet whisper, "Let me give you some advice." 4) The realization that there's nothing but lab to talk about. That's the threshold. 5) The rationalization alpha - The view from 30,000 feet isn't terrible. 6) The rationalization beta - The rundown of broad problems they're having. This is the point where they will probably, as if by magic, remember that thing they were going to do needs to be done now. (I've got some analysis running I need to check, I need to feed some lab animals, I promised my parents I would call, I told a lab mate I'd help them with this thing and will be up all night, etc.) 7) The rationalization gamma - Specific cases of major problems they're seen other have. 8) The rationalization delta - Specific problems they're having.

Re: PhD Simulator

#165
post #102

Earlier quoted context omitted.

> Teaching obligations any advice for people aiming for teaching instead of all the publishing stuff?

Yeah don’t do a PhD

Tried that, the university considered lecturers to be disposable if there was a chance to replace them with a tenure-track who could get grants, and told me I could come back if I got a PhD.

Re: PhD Simulator

#166

Earlier quoted context omitted.

> Teaching obligations any advice for people aiming for teaching instead of all the publishing stuff?

In academia, your Resume/CV is basically a list of what you have published. Even if you are an awesome teacher, you are going to be required to continue publishing a minimum amount every few years and you will be hired based on what you published. Sorry, but that's just academia. If you want to teach without doing research, then maybe look at Community Colleges, High Schools, or getting a job at a corporate job and b…

There are also smaller (typically private) colleges and universities that heavily focus on their undergrad programs. The Jesuits seem to lean into this style with both Santa Clara U. in the Bay Area and Loyola Marymount U. in LA falling into this pile. Research at these institutions definitely ends up taking a secondary role.

Re: PhD Simulator

#167

Earlier quoted context omitted.

This would be incredibly bad advice in half of Physics and most of Math. An adviser would simply not trust a graduate student with middling grades to be competent enough to work with.

I think this depends on the field. For example in CS, my advisor straight up told me multiple times to stop worry about class work. His exact statement is that "There isn't anything more that you will learn in classes that you won't learn in greater detail doing research". His logic is that when you are doing research, you are pushing the envelope into new territory that can't be taught in a classroom. When you are i…

That's not very unique of your advisor. Most researchers are there for the research, not for the teaching and it shows.

Re: PhD Simulator

#168
post #158
post #73

Earlier quoted context omitted.

> You found one of your ideas appears in a recently published paper. You can no longer work on it. This is one of the things I thought of right away when ChatGPT got released last year. "God, there's probably so many PhD candidates right now in NLP feeling despair like all their work was pointless ...as if million of voices cried out in terror and were suddenly silenced." It's hard in the moment to know whether what…

Why is this the case? Wouldn’t having more than one paper proving/discovering the same thing be good for confidence in either of them?

In theory yes, in practice many journals are only interested in work with a clear novelty factor.

Re: PhD Simulator

#169
post #158
post #73

Earlier quoted context omitted.

> You found one of your ideas appears in a recently published paper. You can no longer work on it. This is one of the things I thought of right away when ChatGPT got released last year. "God, there's probably so many PhD candidates right now in NLP feeling despair like all their work was pointless ...as if million of voices cried out in terror and were suddenly silenced." It's hard in the moment to know whether what…

Why is this the case? Wouldn’t having more than one paper proving/discovering the same thing be good for confidence in either of them?

Its sort of a mix of a lot of small things - 1) The coming conferences will be flooded with LLM analysis, so the space will be heavily saturated and more difficult to find a significant contribution; 2) LLMs are a new model that you might need to include in your analysis, which means learning about and becoming familiar with them; 3) your work might get overshadowed because its now obsolete in the land of LLMs

A slight equivalent I can think about would be the emergence of neural networks. When I was working on my Masters on face recognition, neural networks were not the major force they are now. Facial landmarks used a combination of haar features and edge detection. These methods weren't outright abandoned, but if NNs had taken off during my research, then I would have had to restart my work.

Re: PhD Simulator

#170

Earlier quoted context omitted.

doing a PhD for the earning potential is hilarious. you'd be better off getting a normal job, living frugally, and pumping as much into savings for the same amount of time

Same can be said for a startup.

the long tail of profit in a startup is wildly higher than a PhD. To be clear, I say this as one who's gone through a math PhD; none of my fellow graduates make significantly more than they would've made bypassing the PhD for industry, especially when you consider the opportunity costs. Academia is very much for people who either prefer ideas or prestige to money.
Post reply on HN