Should I Get a PhD?
When I was fourteen, I knew that I was destined to become a physics professor, so when I finished my BA at UC Berkeley1 it never even crossed my mind to do anything but go to graduate school. After all, one must first go to graduate school to become a professor, and a professorship was in my future. Ergo, graduate school (specifically, the University of Minnesota2) was the next step in my journey.
Seven years later, as a freshly minted PhD in particle physics, I went straight to Insight Data Science to begin a career in Silicon Valley. The dream of becoming a physics professor had long since been abandoned. I never even applied for a postdoc position, which would have been the next step in the process.3
So what caused me to redirect my ambitions and energy? What was it that, despite thoroughly enjoying my six years in Minnesota,4 has led me to wonder if I would still do it again were I to be given a second chance? And what should my answer be when I am asked by students whether they should pursue a physics PhD? These are not easy questions to answer, nor will any particular answer be suitable for all, but if you are wondering if you should follow in my footsteps (or wondering about the road not taken), then a review of my experiences may reveal insights.
You learn a lot
I learned some highly specialized knowledge in graduate school—relativity, quantum mechanics, and particle cross-sections. Other knowledge was extremely useful, but completely unrelated to physics—how to work with large datasets, how to be skeptical of my conclusions, how to communicate results, and how to manage my time. But the most useful things I learned were about myself—to have confidence in my abilities, that I was a worthwhile human, that I was more than just a physicist.
I understand myself better having gone through grad school, and I am in many ways a more complete person. I also developed some marketable skills, but what I feel like I really gained was the confidence to sell the skills that I did have.
And it is a lot of fun
Graduate school, if you find a good adviser, is also a lot of fun. You meet hundreds of people who share the exact same passion as you, and you get to spend six years having lunch and dinner parties with them all while riding bikes and hiking and rock climbing and whatever else you like to do. I met many of my best friends in graduate school and, with the exception of Insight, it has been the best source of contacts in my professional network.
But there are no jobs
Graduate school was a slow yet constant realization that there were no jobs in my field. Going into grad school, I had naively assumed that everyone, or nearly so, who wanted a professorship got one. But as I watched brilliant postdocs leaving the field one after another, each failing to get even a single offer after trying for years, I realized that I had been wrong. The year I left with my PhD, my experiment of nearly 3000 scientists placed only around ten postdocs out of one-hundred into US-based tenure track positions. Those are terrible odds.
And the opportunity cost is very high
You don’t spend your own money on a graduate degree,5 but it is still very expensive. You spend six or seven years training for a job you will not get. You learn a lot along the way, but you could certainly learn the useful things in a much shorter amount of time.
In six years, if you get rid of the physics courses, exams, and teaching, you could fit a lot of training and on-the-job experience. Most of the experience I acquired during graduate school was not at all applicable to the job I have now, and most hiring managers view it that way as well.
But wait, I want to be a data scientist
It is true that many data scientists (and certainly most of the ones at the companies I’ve worked for) have PhDs, but I think that is an artifact of how new the position is. Teaching the useful skills in a much shorter amount of time than required for a PhD is what the master’s degree programs in data science, and eventually the undergraduate degrees, will attempt to do. And I believe they will succeed after some trial and error.
So if your goal is to be a data scientist, ask yourself this: would you rather have a master’s in data science, and four or five years of experience in that industry, or would you rather be fresh out of a PhD program with no “practical” experience in the field? Several of my friends are in the latter position and I can tell you: it isn’t easy!
Would I do it again?
It’s hard to answer because I really like how my life has ended up. I would be in a completely different place—in terms of my family, my career, and my personality—had I not gone. And yet, for the reasons above, it is hard for me to recommend the same path to others. If you absolutely know you are going to be a professor,6 then you should pursue your goal and enter a PhD program, but if you just want to break into data science, you should consider other options.
Go Bears! 🐻 ↩
A postdoc is a position used to gain more experience before applying for professorships. Graduates often spend five to six years doing multiple postdocs, which are all but required to land a professorship. ↩
I was forced to take a year off after undergrad; a story I shared in another post. ↩
Or, at least, you should not! ↩
You probably aren’t. ↩