So you want to do a Ph.D.

10 February 2020

Here are some things I wish I had known when I started a Ph.D, in no particular order. This post was born as a monologue for undergrads who asked “what is a Ph.D. like?”, and solidified in writing in response to a very bright (and bright-eyed, and bushy-tailed) mentee’s e-mail who was wondering about grad school. It is not meant as a guide - this is just my (n=1) experience.

Hoover Tower at dusk my first month at Stanford

A PhD

A Ph.D. is a job where you get almost unlimited freedom to look at hard-to-impossible problems that are highly interesting but not necessarily profitable, and take a stab at solving them.

The two main downsides of this unlimited freedom is that (i) your salary is 1/3 to 1/4 of what you could be making in industry, and (ii) you are signing up for an engagement that could last anywhere between 4 and 7 years (with outliers of 10+ years), depending mostly on luck, and with a formal payoff that only happens at the very end.

In addition to the freedom, there are three upsides.

First (and least important), you become the world expert on a specific topic.

Second, you are be surrounded by incredibly smart and driven people for the duration of your Ph.D., which is pretty awesome.

Third, you learn how to approach really hard and large problems and take a stab at solving them, and you will also probably learn when to give up on a problem that turns out to be impossible.

What it takes

A Ph.D. is not the natural continuation of your education. It is a job.

In my opinion, the key feature you need to earn Ph.D. student is to be resilient. Not wicked smart - I was certainly not the smartest member of my lab, and being wicked smart doesn’t do you much good when the problem you are solving turns out to be impossible (and a good research problem has at least a 50% chance of being impossible - more on this below).

Picking an advisor: fit and funding

The single most important aspect of your Ph.D. is your advisor. Not the institution you go to, or the topic you pick. Different labs can have completely different cultures, resulting in Ph.D. trajectories that have basically nothing in common with each other. The question of “should I do a Ph.D.” is ill-defined - it all depends on with whom you will be doing it.

My advisor Marco likes to say that, when picking an advisor, there are two things to look at: fit and funding. Let’s unpack this.

Funding

Funding is easy. You don’t want to go with an advisor who does not have a high likelihood of supporting you for the duration of your Ph.D. You may have great personal and professional fit, but you will end up working half your time as a teaching assistant, which is a lot of fun but generally contributes nothing to your research. Sometimes you have to take a gamble on a new professor who just started and does not have a robust pipeline of grants yet. On the other hand, I would think twice before doing a Ph.D. with an older professor who is riding out a couple of existing grants.

Fit

Fit is key, and there are two aspects to it: professional fit and personal fit.

In my opinion, professional fit comes down to being excited about the same broad class of problems. When a professor described what they work on (the specific tools, not “how we are making the world better”), do you think “wow, I want to do this” or “that sounds hard and boring”? I have found that I can typically tell the difference within a few minutes.

For instance, both my advisor and I like problems that are theory-heavy, where you can prove some fundamental property by exploiting the mathematical structure of the problem. The problem should have a clear and immediate applications, but neither of us likes to spend too much time actually testing the applications on hardware. I think we had excellent professional fit (I hope he agrees!).

Crucially, the specific project that you will start your Ph.D. on does not matter much. An interesting research problem has a better than 50% chance of being impossible, which means there is a better than 50% chance that the project you start with will not be your thesis. I went through three projects in my Ph.D.: I spent 2 years on a project that turned out to be impossible and we ran out of money, then 2 years (with a six-month overlap) on a second project that relied on an external collaboration that fizzled out, and 2.5 years (with a 1-year loose overlap) on a third project which ended up being my thesis.

When applying for a Ph.D., I did not contact certain labs because I thought I wanted to do space-related projects, and they focused on aeronautics; my thesis topic ended up being on autonomous cars. (and then I was hired by JPL partly on the strength of my experience with the second project, the only one that did not result in publications - a Ph.D. is really about the journey, not the final product)

Personal fit is the most important aspect of picking an advisor, in my opinion. Your advisor will unironically be the most important person in your life throughout your Ph.D. (my now wife was so jealous of Marco, because I spent a lot more time with him than with her). There is no perfect advisor, but it is imperative that you find an advisor that you mesh well with.

Advisors come in a spectrum from micromanaging to super hands-off. Typically, younger tenure-track faculty are much more hands-on: you will have long research discussions with them, they will ask for frequent updates, and they will stay late at night in the lab with you (one of my favorite memories of grad school is when Marco and I stayed in his office until 10:30 p.m. to try and prove a theorem, with his wife calling every 15 minutes to ask “where are you?”. She may have different feelings about that night). But they may also call you fifteen times on Skype on a Sunday afternoon near a deadline.

Conversely, I know of a very tenured professor who likes to have his students go down a rabbit-hole for six months (even though he believes the student’s approach is hopeless) because sometimes that is how great research happens. Students meet with him every few weeks to months and they are generally very relaxed, but average graduation times are somewhere north of 7 years.

There is no right answer, but you should figure out what kind of advising you respond best to (do you prefer to have a demanding boss or to be left alone?) and pick an advisor that does that.

Also, while there is no right answer, there are wrong answers. There exist advisors who are just not good at their job, or who think a research professor’s job is not to advise, or who just aren’t good people, and they will make you deeply miserable for four to seven years. When I was visiting grad schools, I was warned by a department’s students over beers to “avoid that professor, he will ruin your life”.

The best way to find out who to avoid is to talk to existing students outside the lab. Go out with them for beers and listen carefully - and keep in mind that, if they are staying in a bad lab, there may be a bit of Stockholm Syndrome involved. Academia is a tiny world, and you will hear stories.

The PhD and you

A Ph.D. can be a very demanding experience from a personal perspective. The median first-year Ph.D. student is someone who is used to doing very well academically, who can reliably “get the answer”. Often, a lot of your self-worth is tied up in this, in being the smart guy who can find the answer to tough problems. I was certainly that way.

The problem with that is that interesting research problems are impossible more than 50% of the time; and the only way to find out if a problem is impossible is to try to solve it for a couple of years. If you do not separate your self-worth from being able to “solve any problem” (if you do not learn not to take academic failure personally), you will have a miserable time and perhaps burn out. I wish someone had told me this when I started! It sucks when you find out that a theorem is wrong and that you will not submit to ICRA, or when your paper (which you thought was really good!) is rejected for the second time and the rest of the lab goes to a conference in Chile while you stay home, or when your advisor says “maybe we should not submit this paper of yours, help your colleague instead”. And it is easy to become extremely bitter.

If you go for a Ph.D., it is imperative to treat it like a job: sometimes it works, but if it doesn’t it’s not your fault. But it is a fine line to walk: you have to be driven enough to work on a problem for two years, even if you know that it has a good chance of being a futile effort, and you won’t find out for a while.

One thing that helps is to remember that, even if you do not end up submitting a paper, you will have learned a lot, which is all that matters. Again, I was hired at JPL mainly because of things I learned on a project that resulted in no publications (a huge bust, from an academic perspective).

What is a Ph.D.?

To sum it up:

  • A Ph.D is a job where you get an enormous amount of freedom to pursue problems you find interesting, in exchange for paltry pay and a lot of anguish.
  • Your advisor will determine whether your Ph.D. is great or terrible. Find an advisor that you mesh well with personally and who is excited about problems that you like. Don’t worry about the specific problem you will work on, though.
  • A Ph.D. sucks from a mental health perspective. Try to disentangle your self-worth from your research success - often research does not work out, and that is OK. But don’t get jaded, or you will quit. Maybe be invested in your research but not in the results? It’s hard.

Oh, and read phdcomics.com. It is a documentary.