This is for people interested in optimizing for academic fame (for a given level of talent and effort and other costs). Instead of trying to get a PhD and a job in academia (which is very costly and due to “publish or perish” forces you to work on topics that are currently popular in academia), get a job that leaves you with a lot of free time, or find a way to retire early. Use your free time to search for important problems that are being neglected by academia. When you find one, pick off some of the low-hanging fruit in that area and publish your results somewhere. Then, (A) if you’re impatient for recognition, use your results to make an undeniable impact on the world (see Bitcoin for example), or (B) if you’re patient, move on to another neglected topic and repeat, knowing that in a few years or decades, the neglected topic you found will likely become a hot topic and you’ll be credited for being the first to investigate it.
Instead of trying to get a PhD and a job in academia (which is very costly and due to “publish or perish” forces you to work on topics that are currently popular in academia), get a job that leaves you with a lot of free time, or find a way to retire early.
On the bright side, if we forget the “job in academia” part and just focus on the “PhD” part, a PhD can fit these criteria reasonably well.
Before I justify that, I should acknowledge themanyarticlesarguing, withsomejustice, that a PhD will ruin your life. These articles make fair points, although I notice they have a lot of overlap, mostly concluding that if you get a PhD you’ll spend 6+ years running up masses of debt, with massive teaching loads and no health insurance, worked to death by an ogre as you try to spin literary criticism out of novels analyzed to death decades ago.
The obvious solution: don’t do a PhD in a country where taking 7 years to finish is normal; don’t do a PhD unless someone’s paying you to do it; don’t do a PhD in a department that assigns you endless teaching duties; don’t do a PhD in a country without a universal healthcare system; don’t choose a supervisor who exploits their students; and don’t get a literature PhD.
A “don’t” is less useful than a “do”, so here are some possible “do”s I’d suggest as alternatives:
find PhD programmes where the successful students mostly finish within 4 years (in the UK, 3-4 years is a more typical PhD length than 6-7, but there is variation among universities)
explicitly say on your PhD applications that you can’t afford to do the PhD unless the university waives the tuition fee and offers a stipend (this no doubt reduces your chances of getting a PhD place, but if you’re allergic to debts you want to be selective here)
when you visit prospective departments, ask the professors and current PhD students how much teaching PhD students have to do (in some departments it’s 100% optional, and pays you extra)
do a PhD in the UK, which has a health system where most medical services are free at the point of delivery
try to get an idea of how hard your potential PhD supervisors work their students (don’t just talk to the supervisors themselves — try to talk to their current/former students one-on-one as well)
get a PhD in physics, statistics, accountancy, economics, or something else remunerative and popular with employers
With the usual worries about PhDs out of the way, I turn to Wei_Dai’s concerns. The first is the publish or perish issue. If you’re just doing a PhD, the publish or perish imperative is often weaker than for postdocs & professors. (This again varies with the field and the institution. For example, as I understand things, top-tier US economics PhD students normally publish 3 or 4 serious papers, and basically staple them together for their dissertation. On the other hand, some UK physics students get PhDs without publishing any journal papers at all.) The ultimate hurdle for your work is convincing your supervisor and the handful of external examiners reading your dissertation that it’s worthwhile.
Along the same lines, you don’t necessarily have to work on fashionable topics if you’re getting a PhD. It’s quite possible to work on something boring; it need only be just interesting enough to keep your supervisor on board and satisfy your other examiners. (You’ll probably want a margin of safety, though, in case your work ends up more boring than expected.) A more objective (but still approximate) rule of thumb: your PhD should be interesting enough to be accepted by the same rank of journal as the papers it’s citing. If your PhD doesn’t need to serve as a step up into an academic job, it can be as boring as you like as long as it meets the baseline.
Lastly, what about free time? A lot of PhDs eat virtually all of your attention, but some offer ample free time in the first couple of years if the work involved isn’t fiddly. For example, you might end up running lots of simulations with a computer program that’s already been written. If so, you might well be able to go to your office in the morning, set a run going, and spend the afternoon doing something else.
One catch is that it’s not trivial to tell which PhDs are low-effort before the fact. Even if your supervisor accurately tells you what they expect from you, and the other students accurately report that they don’t spend much time poring over their work, you might still get unlucky and end up slaving over a computer or an experiment or some equations for 16 hours a day, because research is unpredictable. (Still, compare it to the main alternative: people routinely underestimate how long they’ll spend at the workplace — and commuting! — for normal jobs, too. It’s not obvious that PhDs are more unpredictable in this regard.)
Nonetheless, if you plan ahead to do straightforward work for an easy-going supervisor who’s not in the office most days, you might well be able to spend most days off campus yourself, doing your own independent research instead. And while you’re a student, there’s nothing stopping you from visiting other departments at your university to pick the brains over there!
Use your free time to search for important problems that are being neglected by academia. When you find one, pick off some of the low-hanging fruit in that area
“don’t do a PhD in a country without a universal healthcare system”
Funded PhD’s in the US commonly include health insurance coverage as part of your stipend.
This is yet more support for your main point: the fact that getting a PhD in some programs/fields is a bad idea does not mean you should avoid a PhD from any program/field.
FWIK, some universities allow you to get PhD in computer science by submitting PhD thesis for review and paying some amount of money (~$1200 on my university). This way, one can follow your advice and still get PhD.
Tell us more. Much more, in excruciating detail. I am reasonably sure I remember reading Eliezer write about the impossiblity of what you just described, i.e. getting a Ph.D. without necessarily having an advisor, funding or a Bachelor’s degree.
A PhD is only as good as the reputation of your advisor. If everybody knows your advisor then you won’t have a problem finding a job in academia. If your PhD is not backed by a prominent professor with a name, you’re going to have a very difficult time finding a good position. It may be a bit easier in CS, where universities have to compete with industry, compared to my field (physics/chemistry/materials science), but generally this is how academia works.
An easily obtainable PhD is generally not the right kind of signal.
A PhD is only as good as the reputation of your advisor. If everybody knows your advisor then you won’t have a problem finding a job in academia.
I would amend this to be “if everybody knows your advisor you’ll have FEWER problems finding a job in academia.” Some fields are very, very crowded (theoretical physics, for instance). For a very brief time, I was in a small team at a consulting company where 3 out of the 4 of us had done a science phd under a Nobel winner, and still ended up making major career transitions after half a decade of postdocs. Science is crowded, the more basic the research the more crowded the field. To first order, no one gets a job. If you are under a famous advisor you might move your odds up to 1⁄10 or 1⁄5 or something like that.
It could be that prestigious second-movers deserve the credit if they are responsible for getting people to pay attention to the previously neglected topics, and possibly we already credit first-movers more than we should (which is why I said “optimize for academic fame” instead of “positive social impact”). Which brings up a question: what determines the topics that academia pays attention to? If we had a good model for that, maybe we could use it to generate some munchkin ideas for making it pay attention to important but neglected ideas?
He has another post about how if you say something outrageous that later becomes common wisdom, you won’t be widely admired for having said it first; you will still be thought of as a crank.
Cognitive bias is now much more popular and fashionable than it was when I first started talking to my friends about it after reading Eliezer’s posts. I predict that zero people will say “so it looks like this Eliezer guy you keep talking about was ahead of the curve on cognitive bias, maybe it’s worth hearing some of his other ideas?”
This setting seems more optimal for actually doing theoretical work of your own choosing without getting distracted by a need to compete or justify your interests. It seems less risky/way easier than trying to get the same benefits while working within academia, but you won’t get the external motivation/guidance/sanity-checking and by default won’t be as close to the professional community.
It seems less risky/way easier than trying to get the same benefits while working within academia
Right, you can get the same benefits in academia by getting tenure, but how many people manage that, and even if you do, the most productive period of your life might already be over by then.
but you won’t get the external motivation/guidance/sanity-checking
This is an important consideration. The external motivation/guidance/sanity-checking provided by the relevant online non-professional communities were enough for me to be productive and not become a crank, etc., but maybe (as cousin_it suggests) I’m very unusual in that regard.
That plan worked for you, but you’re very unusual.
Weren’t you essentially following my plan too, i.e., working in your free time on topics being neglected by academia? (Are you still doing this, BTW, after you quit from being an SI research associate?) Are you implying that the plan isn’t working for you?
You’d probably be an even bigger intellectual celebrity if you took the academic path.
I’m not sure how you figured that. If I had gone into academia I most likely would have gone into computer science and specialized in something not particularly Earth-shattering like crypto optimization (i.e., making crypto algorithms faster), or if I was lucky maybe I could have pursued my b-money idea. But I never would have had the opportunity to pursue my interests in philosophy (which seems to have a chance of making me more famous in the future when academia or posthumans discover or reinvent UDT).
Even if I had somehow gotten a job in academic philosophical research, it took me 3-4 years exploring various dead ends before getting the idea that the solution to anthropic reasoning / indexical uncertainty is in the shape of a decision theory, and then even more years to formulate it into the form you saw in my LW post. I don’t know how I would have survived in academia for those years without any publishable results. Instead what probably would have happened (and what apparently happened to every professional philosopher who actually worked on the topic) is that I would have been forced to quickly come up with some sort of wrong solution just to have something to publish.
This reminds me of the story of Robert Edgar, who created the DNA and protein sequence alignment program MUSCLE.
He got a PhD in physics, but considers that a mistake. He did his bioinformatics work after selling a company and having free time. The bioinformatics work was notable enough that it’s how I know of him.
“Instead of trying to get a PhD and a job in academia (which is very costly and due to “publish or perish” forces you to work on topics that are currently popular in academia), get a job that leaves you with a lot of free time”
Part of the attraction of academia to me is that it is exactly the job that leaves you with lots of free time. A professor only has to be in a certain place at a certain time 3-12 hours per week (depending on teaching load), 30 weeks per year. After tenure, you can research whatever you want, especially if you aren’t in a lab-science field that leaves you dependent on grants. Even before tenure I can work on neglected problems, so long as they aren’t neglected due to their low prestige.
This is for people interested in optimizing for academic fame (for a given level of talent and effort and other costs). Instead of trying to get a PhD and a job in academia (which is very costly and due to “publish or perish” forces you to work on topics that are currently popular in academia), get a job that leaves you with a lot of free time, or find a way to retire early. Use your free time to search for important problems that are being neglected by academia. When you find one, pick off some of the low-hanging fruit in that area and publish your results somewhere. Then, (A) if you’re impatient for recognition, use your results to make an undeniable impact on the world (see Bitcoin for example), or (B) if you’re patient, move on to another neglected topic and repeat, knowing that in a few years or decades, the neglected topic you found will likely become a hot topic and you’ll be credited for being the first to investigate it.
On the bright side, if we forget the “job in academia” part and just focus on the “PhD” part, a PhD can fit these criteria reasonably well.
Before I justify that, I should acknowledge the many articles arguing, with some justice, that a PhD will ruin your life. These articles make fair points, although I notice they have a lot of overlap, mostly concluding that if you get a PhD you’ll spend 6+ years running up masses of debt, with massive teaching loads and no health insurance, worked to death by an ogre as you try to spin literary criticism out of novels analyzed to death decades ago.
The obvious solution: don’t do a PhD in a country where taking 7 years to finish is normal; don’t do a PhD unless someone’s paying you to do it; don’t do a PhD in a department that assigns you endless teaching duties; don’t do a PhD in a country without a universal healthcare system; don’t choose a supervisor who exploits their students; and don’t get a literature PhD.
A “don’t” is less useful than a “do”, so here are some possible “do”s I’d suggest as alternatives:
find PhD programmes where the successful students mostly finish within 4 years (in the UK, 3-4 years is a more typical PhD length than 6-7, but there is variation among universities)
explicitly say on your PhD applications that you can’t afford to do the PhD unless the university waives the tuition fee and offers a stipend (this no doubt reduces your chances of getting a PhD place, but if you’re allergic to debts you want to be selective here)
when you visit prospective departments, ask the professors and current PhD students how much teaching PhD students have to do (in some departments it’s 100% optional, and pays you extra)
do a PhD in the UK, which has a health system where most medical services are free at the point of delivery
try to get an idea of how hard your potential PhD supervisors work their students (don’t just talk to the supervisors themselves — try to talk to their current/former students one-on-one as well)
get a PhD in physics, statistics, accountancy, economics, or something else remunerative and popular with employers
With the usual worries about PhDs out of the way, I turn to Wei_Dai’s concerns. The first is the publish or perish issue. If you’re just doing a PhD, the publish or perish imperative is often weaker than for postdocs & professors. (This again varies with the field and the institution. For example, as I understand things, top-tier US economics PhD students normally publish 3 or 4 serious papers, and basically staple them together for their dissertation. On the other hand, some UK physics students get PhDs without publishing any journal papers at all.) The ultimate hurdle for your work is convincing your supervisor and the handful of external examiners reading your dissertation that it’s worthwhile.
Along the same lines, you don’t necessarily have to work on fashionable topics if you’re getting a PhD. It’s quite possible to work on something boring; it need only be just interesting enough to keep your supervisor on board and satisfy your other examiners. (You’ll probably want a margin of safety, though, in case your work ends up more boring than expected.) A more objective (but still approximate) rule of thumb: your PhD should be interesting enough to be accepted by the same rank of journal as the papers it’s citing. If your PhD doesn’t need to serve as a step up into an academic job, it can be as boring as you like as long as it meets the baseline.
Lastly, what about free time? A lot of PhDs eat virtually all of your attention, but some offer ample free time in the first couple of years if the work involved isn’t fiddly. For example, you might end up running lots of simulations with a computer program that’s already been written. If so, you might well be able to go to your office in the morning, set a run going, and spend the afternoon doing something else.
One catch is that it’s not trivial to tell which PhDs are low-effort before the fact. Even if your supervisor accurately tells you what they expect from you, and the other students accurately report that they don’t spend much time poring over their work, you might still get unlucky and end up slaving over a computer or an experiment or some equations for 16 hours a day, because research is unpredictable. (Still, compare it to the main alternative: people routinely underestimate how long they’ll spend at the workplace — and commuting! — for normal jobs, too. It’s not obvious that PhDs are more unpredictable in this regard.)
Nonetheless, if you plan ahead to do straightforward work for an easy-going supervisor who’s not in the office most days, you might well be able to spend most days off campus yourself, doing your own independent research instead. And while you’re a student, there’s nothing stopping you from visiting other departments at your university to pick the brains over there!
I don’t have any tips for this, though.
“don’t do a PhD in a country without a universal healthcare system” Funded PhD’s in the US commonly include health insurance coverage as part of your stipend.
This is yet more support for your main point: the fact that getting a PhD in some programs/fields is a bad idea does not mean you should avoid a PhD from any program/field.
FWIK, some universities allow you to get PhD in computer science by submitting PhD thesis for review and paying some amount of money (~$1200 on my university). This way, one can follow your advice and still get PhD.
Tell us more. Much more, in excruciating detail. I am reasonably sure I remember reading Eliezer write about the impossiblity of what you just described, i.e. getting a Ph.D. without necessarily having an advisor, funding or a Bachelor’s degree.
Seconded.
A PhD is only as good as the reputation of your advisor. If everybody knows your advisor then you won’t have a problem finding a job in academia. If your PhD is not backed by a prominent professor with a name, you’re going to have a very difficult time finding a good position. It may be a bit easier in CS, where universities have to compete with industry, compared to my field (physics/chemistry/materials science), but generally this is how academia works.
An easily obtainable PhD is generally not the right kind of signal.
I would amend this to be “if everybody knows your advisor you’ll have FEWER problems finding a job in academia.” Some fields are very, very crowded (theoretical physics, for instance). For a very brief time, I was in a small team at a consulting company where 3 out of the 4 of us had done a science phd under a Nobel winner, and still ended up making major career transitions after half a decade of postdocs. Science is crowded, the more basic the research the more crowded the field. To first order, no one gets a job. If you are under a famous advisor you might move your odds up to 1⁄10 or 1⁄5 or something like that.
email me with info about that company, OK?
Sounds like maybe MetaMed should inquire into working with them.
Extraordinary claims....
Hanson has a post somewhere about how the first-movers often don’t get credited, just the prestigious second-movers.
Sociology of science calls this the Matthew Effect
Ohh. “Kolmogorov Complexity” was actually invented by Solomonoff. Interesting.
It could be that prestigious second-movers deserve the credit if they are responsible for getting people to pay attention to the previously neglected topics, and possibly we already credit first-movers more than we should (which is why I said “optimize for academic fame” instead of “positive social impact”). Which brings up a question: what determines the topics that academia pays attention to? If we had a good model for that, maybe we could use it to generate some munchkin ideas for making it pay attention to important but neglected ideas?
I hope the irony was intentional. (Here’s the post, btw.)
He has another post about how if you say something outrageous that later becomes common wisdom, you won’t be widely admired for having said it first; you will still be thought of as a crank.
Cognitive bias is now much more popular and fashionable than it was when I first started talking to my friends about it after reading Eliezer’s posts. I predict that zero people will say “so it looks like this Eliezer guy you keep talking about was ahead of the curve on cognitive bias, maybe it’s worth hearing some of his other ideas?”
This setting seems more optimal for actually doing theoretical work of your own choosing without getting distracted by a need to compete or justify your interests. It seems less risky/way easier than trying to get the same benefits while working within academia, but you won’t get the external motivation/guidance/sanity-checking and by default won’t be as close to the professional community.
Right, you can get the same benefits in academia by getting tenure, but how many people manage that, and even if you do, the most productive period of your life might already be over by then.
This is an important consideration. The external motivation/guidance/sanity-checking provided by the relevant online non-professional communities were enough for me to be productive and not become a crank, etc., but maybe (as cousin_it suggests) I’m very unusual in that regard.
That plan worked for you, but you’re very unusual. You’d probably be an even bigger intellectual celebrity if you took the academic path.
Someone closer to average, like me, cannot do research alone, only in a group of like-minded people.
Weren’t you essentially following my plan too, i.e., working in your free time on topics being neglected by academia? (Are you still doing this, BTW, after you quit from being an SI research associate?) Are you implying that the plan isn’t working for you?
I’m not sure how you figured that. If I had gone into academia I most likely would have gone into computer science and specialized in something not particularly Earth-shattering like crypto optimization (i.e., making crypto algorithms faster), or if I was lucky maybe I could have pursued my b-money idea. But I never would have had the opportunity to pursue my interests in philosophy (which seems to have a chance of making me more famous in the future when academia or posthumans discover or reinvent UDT).
Even if I had somehow gotten a job in academic philosophical research, it took me 3-4 years exploring various dead ends before getting the idea that the solution to anthropic reasoning / indexical uncertainty is in the shape of a decision theory, and then even more years to formulate it into the form you saw in my LW post. I don’t know how I would have survived in academia for those years without any publishable results. Instead what probably would have happened (and what apparently happened to every professional philosopher who actually worked on the topic) is that I would have been forced to quickly come up with some sort of wrong solution just to have something to publish.
This reminds me of the story of Robert Edgar, who created the DNA and protein sequence alignment program MUSCLE.
He got a PhD in physics, but considers that a mistake. He did his bioinformatics work after selling a company and having free time. The bioinformatics work was notable enough that it’s how I know of him.
His blog post, from which I learned this story: https://thewinnower.com/discussions/an-unemployed-gentleman-scholar
“Instead of trying to get a PhD and a job in academia (which is very costly and due to “publish or perish” forces you to work on topics that are currently popular in academia), get a job that leaves you with a lot of free time” Part of the attraction of academia to me is that it is exactly the job that leaves you with lots of free time. A professor only has to be in a certain place at a certain time 3-12 hours per week (depending on teaching load), 30 weeks per year. After tenure, you can research whatever you want, especially if you aren’t in a lab-science field that leaves you dependent on grants. Even before tenure I can work on neglected problems, so long as they aren’t neglected due to their low prestige.
Yes but before you get tenure you’ve wasted your most productive and fun youthful years getting tenure.