I've posted a few things seeking career advice with mixed success. In this case I have a more concrete question and if you feel like commenting, I'd appreciate it. I think it helps me to hear what a community of others thinks from a rational perspective because there are often many components to a decision that I had not anticipated.

I am currently a grad student working in computer vision. I dislike the way that my current adviser focuses only on projects that have short-term commercial gains. I want to study more fundamental, theoretical research which may take more time to develop but will also be more aesthetically pleasing to me. For me, the only reason to agree to be paid so little as a graduate student is to gain the opportunity to work freely on high risk projects that happen to be of personal interest. Practical considerations are not interesting to me as motivation for a Ph.D. On the other hand, it has felt nearly impossible to actually find faculty willing to have students work on theory. Rather than grinding away with no dental insurance for 3 more years, followed by low paying post-docs, etc., perhaps seeking a job will be better.

I have some interesting job prospects that are all with larger companies. The jobs are basically business analytics, including scientific computing, data mining, and machine learning. I'm sure the problems to work on are not that great; not going to be Earth shattering, but at the same time they sound a lot more interesting to me than hedge fund data analysis or military research labs (I have working experience at a government lab and I did not enjoy it). The hours would be better; the pay is fair and it would be a good living. I could pursue some things as serious hobbies outside work.

At the same time though, there feels like a nagging opportunity cost. I am not naive enough to believe there will be a nice faculty job waiting for me even if I finish my Ph.D. However, I really enjoy theoretical and mathematical physics, machine learning, computational complexity, and scientific computing, and various philosophical considerations generated by these. Being able to teach about them, research them, and work on them professionally seems incredibly appealing. Am I making a big mistake if I leave? How can one pursue philosophical interests and desires to work in theory outside of a typical job? Or should I even worry about such a thing?


New Comment
27 comments, sorted by Click to highlight new comments since: Today at 6:33 PM

As far as I can tell you identify two options: 1) continue doing the PhD you don't really enjoy 2) get a job you won't really enjoy.

Surely you have more options!

3) You can just do a PhD in theoretical computer vision at a different university.

4) You can work 2 days a week at a company and do your research at home for the remaining 4 days

5) Become unemployed and focus on your research full time

6) Save some money and then move to Asia, South America or any other place with very low cost of living so you can do a few years of research full time.

7) Join a startup company that is doing groundbreaking computer vision work

8) See if there is something else that you can be passionate about and do that.

Life's too short to do something you don't enjoy and you're now at a point in your life where the decisions you make are going to have real consequences. So do some soul searching and figure out what you really want and then figure out what you have to do to make it happen. That's life 101.

When you're spending the majority of your time doing something you don't really enjoy you have a big problem. This is the only life you have and it's easy to waste it 5 years at the time! Maybe your true dream is to work on the next Pixar movie, or to design special effects for the next CGI blockbuster! But if you aren't going to explore your options seriously you're not going to find out what you really want to do in life. If, on the other hand, you're absolutely sure you want to do theoretical computer vision research, then JUST DO THAT. There are thousands of universities with good computer vision departments. So unless you got a 1000 rejection letters on your desk you haven't even seriously explored your options yet.

(PS: Forget about doing research in the evenings after you get home from a day job. It doesn't work. Many people do this and then they figure out that after a full day's work you don't have the energy anymore to do really difficult stuff. Your lifestyle will change and you'll grow dependent on your job. Then as you get older you'll look back and call it a "silly dream" and wisely observe that you have to make compromises in life and that your ability to compromise on what you want makes you a responsible adult.)

(PPS: I'm trying to convey that being unhappy with your job should trigger "hair on fire" like panic.)


I have already transferred schools once, moving because there were no advisers in my area at school #1 (the one I had planned to work with became emeritus right as I joined). I like the school I am at now a lot more than I like computer vision. In fact, my main issue with my current situation is that it appears that no one can do fundamental research in computer vision: all of the major conferences require you to pander to shorter term commercial applications if you want to publish and I'd rather move to a new field than jump through those hoops.

I don't consider options 4, 5, 6, or 7 to be remotely realistic for me. I can't think of an Asian or South American countries to where I would be happy with the governments or the long distance from family and friends if I were to live there semi-permanently. Those considerations are at least as important to me as job considerations. I don't consider unemployment or extreme part-time work an option because I have other life goals, like traveling, home ownership, etc., that I want to financially support in addition to whatever career path I choose.

I appreciate your suggestions, but I have really thought about this a considerable amount. The post that I linked above has some more details about what thinking I have already done. I would really appreciate more targeted advice if you are interested. Given the climate for faculty jobs, what is the best way to try to achieve one? What are ways to do theoretical work / teaching at a university level for a living that are non-traditional?

When you say "I have really thought about this a considerable amount", I hear "I have diagnosed the problem quite a while ago and it's creating a pit in my stomach but I haven't taken any action yet". I can't give you any points for that.

When you're dealing with a difficult problem and if you're an introspective person it's easy to get stuck in a loop where you keep going through the same sorts of thoughts. You realize you're not making much progress but the problem remains so you feel obligated to think about it some more. You should think more, right? It's an important decision after all?

Nope. Thinking about the problem is not a terminal goal. Thinking is only useful insofar it leads to action. And if your thinking to action ratio is bad, you'll get mentally exhausted and you'll have nothing to show for it. It leads to paralysis where all you do is think and think and think.

If you want to make progress you have to find a way to decompose your problem into actionable parts. Not only will action make you feel better, it's also going to lead to unexplored territory.

So what kind of actions can you take?

Well, your claim is that major conferences require short term commercial papers. So if you go systematically through the papers published in the last year or so you'll find either (a) all the papers are boring, stupid, silly or wrong. (b) there are a bunch of really cool papers in there. In case of (a) maybe you're in the wrong field of research. Maybe you should go into algorithms or formal semantics. In this case look at other computer science papers until you find papers that do excite you. In case of (b) contact the authors of the papers; check out their departments; etc, etc.

To recap: Find interesting papers. Find departments where those interesting papers were written. Contact those departments.

Another strategy. Go to the department library and browse through random books that catch your eye. This is guaranteed to give you inspiration.

This is just from the top of my head. But whatever you do, make sure that you don't just get stuck in a circle of self-destructive thought. Action is key.

If you're certain you want to eventually get a faculty job, do a combination of teaching and research, own a house and regularly go on holiday, then I can't think of any alternatives to the conventional PhD -> faculty route. What's the best way to achieve a faculty job? I don't know. Probably a combination of networking, people skills and doing great research. If you want a faculty job badly enough you can get one. But once you get it there's no guarantee you're going to be happy if what you really want is complete autonomy.

I'm sorry I can't give any targeted advice.

(PS: some people like the idea of travel more than they like travel and some people like the idea of home-ownership more than they like home-ownership. For instance, if you haven't traveled a lot in the past 5 years you probably don't find travel all that important (otherwise you would've found a way to travel).)


If you want a faculty job badly enough you can get one.

I disagree. I think much of the evidence about the rise of post-docs as principal investigators and the diminishing number of tenured positions is at odds with this claim. This claim is essentially why most students go to a Ph.D. program and they become depressed when they learn it doesn't work like this about 3 years into the process.

For instance, if you haven't traveled a lot in the past 5 years you probably don't find travel all that important (otherwise you would've found a way to travel).

In the last 5 years, I've taken low-paying math research jobs several summers so that I could live in Paris, Hong Kong, and College Station, TX, just to experience parts of the world I had not been to. I've moved (at great personal expense) 3 times in the past 4 years to get out of life situations that I found unsatisfactory. I think that my thinking-to-action ratio is not bad.

You seem to dismiss the possibility that there can be real life Catch-22s. Given my preferences, I think I am in a Catch-22 and I cannot determine an actionable step. Some of my favorite life advice came from a high school math teacher who said "when you don't know what to do, do something." I think I am more insightful than just to wallow in akrasia.

So if you go systematically through the papers published in the last year or so you'll find either (a) all the papers are boring, stupid, silly or wrong.

Yes, this is exactly what I have been doing for the past 2 years. But when I have discussed the option to switch to other research fields with faculty and older graduate students, they are telling me that the condition (a) is going to be true in every research field where there is actually enough grant money to finance my studentship, and that (a) is just a part of life in science and that I should be more focused on just doing programming tasks and coming up with small software developments that cater to commercial interests, leading to papers that fit into condition (a). I completely reject their point of view; I think they are wrong, and I think that if academia is set up this way, then my options are to leave academia for jobs that I think are very suboptimal or else agree to unhappily suffer through the academic hoop-jumping that I don't like.

Given that these are my only options, I am trying to prepare myself to choose one or the other. But the biggest opportunity cost that I feel scared about is losing the chance for theoretical research and philosophical aspects of science to be a major component of the value that I contribute over my career.

It seems you have added a lot of constraints to your problem, consider that you can add only so many until it becomes impossible to solve.

I don't get what is so bad about the part-time job if you wrote that:

Rather than grinding away with no dental insurance for 3 more years, followed by low paying post-docs, etc., perhaps seeking a job will be better.

So you aren't making much at the university either are you?


Yes, there are a lot of constraints. They aren't all hard constraints, but that's life. I'm not going to put all my eggs into one basket, but I think I would be about as significantly unhappy if I can't resolve career issues as if I can't live within a 4-hour plane ride radius from my family.

Regarding the part-time job thing: it would be pretty hard to sustain myself financially that way. Like I said, I have additional goals like saving money for retirement/home ownership. I don't see this as being orthogonal to doing a career I derive personal meaning from at all: being able to do this is a standard byproduct from most careers at the earnings levels that my master's degree would provide. How do I pinpoint the subset of those careers that also satisfy the constraint that they pay me money to do mathematical research?

I like the school I am at now a lot more than I like computer vision.

Are your computer vision skills transferable to other fields that you can do your preferred type of research in at your current university?

Are there related fields that you think you would be interested in, and would be willing to branch out into?

6) Save some money and then move to Asia, South America or any other place with very low cost of living so you can do a few years of research full time.

I am fascinated by this idea in principle, but do you know anyone who has actually done it? I fear there are many nonobvious details that would derail the plan. Maybe we should create an LW outpost in Saigon or Bangalore or some other inexpensive place, since there are many people here who are excited about the idea of living inexpensively to free up more time for Big Thinking.

I know several people who moved to Asia to work on their internet startup. I know somebody who went to Asia for a few months to rewrite the manuscript of a book. In both cases the change of scenery (for inspiration) and low cost of living made it very compelling. Not quite the same as Big Thinking, but it's close.



I'm flattered, but I'm only occasionally coherent.

No, I concur with GabrielDuquette.

And by the way, I, um, have a friend in the position you've described in your reply there. What's the general template for, um, him, to get out of that situation?


"You must concentrate upon and consecrate yourself wholly to each day, as though a fire were raging in your hair." - Taisen Deshimaru


Useful in other contexts, but not when the problem is choosing direction in your life over a longer term than one day.

[This comment is no longer endorsed by its author]Reply

Great answer!

I want to emphasize the following:

4) You can work 2 days a week at a company and do your research at home for the remaining 4 days

You could start freelancing(there are sites like http://www.vworker.com) and work as much as you want/need to live comfortably. If you have a lot of knowledge you can make good money consulting. I think you can make enough for a living if you work at most 30 hours per week(this would be 3 days x 10 hours). That gives you another 4 days of free time per week to focus on whatever you want to do.




For a consult, x-rays, a cleaning, and 3 fillings in the greater Boston area, it was more than $750 dollars, and in the last calendar year I also required a crown and I am still paying off the debt on that. In total it was more than $1200. That may not seem like a lot, but on a graduate student stipend, that is a lot even spread out over 12+ months. In general, though, I don't have access to funds that would cover dental emergencies. I try to take good care of my teeth, but having a fall back plan of "I'll just pay out of pocket at a cheap dentist" seems bad to me.



I understand your concern, but it may unrealistic to have your security cake and eat it too. Do you have a reason to believe dental emergencies will occur in the near future?

Well, given the amount of cake he's eating...


Isn't one of the main points of insurance specifically to have financial security for things you can't predict? Hedging against health risk? I don't think I will be abnormally prone to dental problems in the next year, but we're also talking about lack of dental insurance for the next 3-4 years, over which it is more likely that I will need things that would massively cheaper with insurance than without. I also don't have much of a lifeline with extended family; they couldn't really afford to help much.



I'd claim that this had nothing to do with rationality, but I know that people would tell me that it was "Rationality applied to finding work."


I'm trying to avoid akrasia and decide how to change preferences to match what I perceive as a catch-22 in terms of job choices. Please see the other post for some background. I agree the relationship between my post and many of LW's goals is tenuous, but I feel like asking a community of rationalists for advice is a rational thing to do and that the signal-to-noise ratio of the discussion board is such that my post is not problematic or inappropriate.

You also can finish your PhD and then get an industry job. Your field has plenty of research scientist jobs. Some of them would be really cutting edge, too.

Also, if your degree is in CS, you should not need a postdoc either, before applying for faculty jobs. If it's EE/CE, I don't know. If you really want to do theoretical stuff, it pretty much requires a faculty position, perhaps at a small liberal arts college where they don't care about grants much. Such places exist, and even have good (undergraduate) students.


There are a few issues with this. First, it is not at all true that CS degree holders can skip the post-doc. There is a lot of evidence that post-docs are rising as principal investigators in research grants and that, as most universities reduce the number of tenured positions available, they are supplementing the workload with mostly post-doc positions because they can be paid less and are entitled to fewer benefits. In fact, when I worked at MIT Lincoln Laboratory, it was actually an openly discussed part of their business model that hiring people as post-docs was more cost effective since they could hire 2 or 3 post-docs for the same price as a single full-hire at the PhD level. Because their government projects had a turnaround time of about 2-3 years, it was a perfect fit and even if the post-docs left the lab, there was not any serious corporate brain-drain going on. A few permanent engineers plus a constantly churning team of 20 post-docs could easily crunch through years worth of government projects without losing much to the high employee turnover. From what I understand, Los Alamos and Sandia National Labs are also adopting this business model. Essentially, positions are offered as being "prestigious post-doctoral fellowships" that pay ~40k/year for a job that most other employers would have to pay 70-100k/year for. I am even starting to see this in private corporations, such as Kodak and IBM where they also have rotating systems of post-docs instead of hiring long-term employees. One part of this is just that with the rise of certain technologies and with an increasingly risk-averse stance toward basic science research, the wages for skilled scientific workers are just plummiting.

The other thing, which is mentioned in this Economist article, is that in mathematics and computer science, someone who leaves a PhD program early with only a master's degree will earn exactly the same wage as someone with a PhD, on average. Coupled with the fact that the master's degree holder will have a 2-4 year lead time working with a company, they will be closer to a promotion, have more years of experience, and more years of retirement savings. Over the lifetime of a career, stopping with only a master's degree will allow you to earn more money than if you finish the PhD, and this is across all work sectors.

So basically, my dilemma is this: if I am just going to be forced to take one of these industry kind of research jobs, then it is economically more rational for me to just stop with a master's degree and begin working now. The number of hours I work as a student is severe, so taking a regular 9-5 job would lower my stress level, give me health insurance, increase my financial situation, and allow me some amount of time to pursue hobbies outside of work. Currently, in school, all of those areas are bad for me.

After working for two years as an algorithm developer for an air force lab, I really dislike the military/government research sector and I want to avoid working there if possible. I have some small interest in financial mathematics, and could probably get a job at a hedge fund, but I am not sure I will really be happy in that stressful situation. Having money in the short term has never been a truly significant motivator for me. Having lots of money in the long term is more motivating because I could then be an investor who finances PhD studentships myself and potentially I could commission more work to be done in the basic science research that I think is being ignored across academia today. But is it worth working doggishly through my young life to potentially be a wealthy investor later on? I don't know.

This leaves me feeling like my only real industry options would be for corporations, some of which are just as bad as government labs. I have found some interesting options in the entertainment / web development domain that, even though they do not involve mathematical theory or deeply intellectual pursuits, do combine machine learning, scientific computing, and seem to be places where a skilled mathematician could succeed in ways that typical workers for these jobs could not.

I'm just trying to decide whether committing myself to jump ship out of a PhD is really worth it if one of my main long-term preferences is to "contribute to basic science research." Really, my goal might be better stated as "become the Radiohead of science" and start my own version of something like github, but instead of software development, it would be for farming out small actionable steps in larger scientific research projects. Do I really, truly believe this is something I can do in my spare time while pursuing career success for an entertainment company, software firm, etc.?


I was doing some research for a small piece I am writing on chess heuristics and human abstraction and came across this video of Vishy Anand. It was inspiring and echoes some of Zed's advice, at least the parts that are relevant to my situation right now.


This post from OB might be somewhat relevant.

New to LessWrong?