from Guide to the PhD on Oct 30, 2022
Why pursue a PhD? Is it for me?
Many reasons for and against doing a PhD are based on oversimplified rumors, and context is the best remedy. Let me clarify what you heard on the grapevine.
If you're uncertain about graduate school ambitions, specifically for a PhD, this post is for you. Once upon a time, I was absolutely certain I would bail from the school at the earliest possible opportunity — a masters wasn't even on my radar, much less a PhD. Yet, here I am today. I've started and heck, even finished a PhD.
However, in retrospect I realize that many would-be researchers chose not to apply based on the wrong information; yet others left the PhD disappointed for the same reason. To save you from both fates, let me clarify several misconceptions about the PhD, reasons both for and against doing one.
This post is written from a 30,000-foot perspective. It's not meant to be a list of criteria to make decisions with, and instead, the post is designed to add context to rumors. This post should clarify confusion but might augment any indecisiveness, putting you right back on the fence. Regardless, I hope that this post makes you ultimately more confident in your decision, whichever way you swing.
There are roughly three categories of potential audiences; regardless of which audience you belong to, this post is written for you, and separate from this post, there are clear action items that make this post relevant to read:
- Anyone without research experience - ideally an early undergraduate in freshman or sophomore year. At this point, you should try research regardless of your academic or industry inclination. Specifically at UC Berkeley, there are plenty of research opportunities abound, and it's a fairly established part of the undergraduate experience. This post can additionally help you understand if a PhD is worth entertaining. To find research opportunities, see How to get into research, as an undergraduate.
- Anyone with research experience - likely an undergraduate in junior or senior year. Bluntly said, you don't have a decision to make until you get offers. It's possible hiring or admissions committees will make the decision for you, so you should apply no matter what — then decide later. In the meantime, before application season, this post can help you understand how earnestly you should pursue research. For the application itself, see How to write your personal statement, for PhD admissions.
- Applicant with offers in hand and deciding between academia and industry. This post can help you understand which decision criteria are based off of faulty hearsay. I hope you've heard of this clarifications before, but if not, these are must-knows before choosing.
If you're reading this blog post, you're probably familiar with the basics. Regardless, I'll go over a few clarifications to address common misconceptions:
- You don't pay tuition; you get paid. This is different from a masters or other advanced degrees like medical school. The stipend is small, sure, but positive cash flow is better than negative.
- You aren't barred from internships or forced to teach. It all depends on your adviser. If they want you focused on research, they may stop you from pursuing an internship. If they are low on funds, they may ask you to take a teaching position. However, I've been fortunate to have an adviser that did neither of these things, so this will depend almost entirely on your adviser.
That's it for the basics. Below, I'll challenge several ideas you may have heard. These opinions below will certainly sow disagreement with someone else, but they're opinions I stand by pretty strongly, after my own experiences and discussions with my peers. These are bad reasons to pursue a PhD, and right after, we'll cover bad reasons to do a PhD1.
There are a number of good reasons to pursue a PhD, but there are also a slew of misconceptions that accompany those good reasons. If you find one of your key motivations for graduate school below, it's worth reconsidering:
❌ I want to be "in" the inner circle. I want to AI. You can also replace "AI" with whichever hype field you want. Don't do a PhD to be "in," because pursuing academia won't help. PhD students, postdocs, and even professors sometimes don't feel "in". The feeling of missing out doesn't ever really go away. To understand why, consider how quickly the field moves:
- Arxiv—an informal hub for paper uploads—features 1200 daily uploads. It would take you effectively the whole day to simply read abstracts. To help with the information overload, several famous content creators scour these streams and curate lists of interesting papers. Unfortunately, there are also a large number of content creators, so even with these curated lists, there are still many more papers than time.
- There's clear evidence that this fast pace impacts academics in the field too. If you're not convinced that this has a deleterious effect, see papers at the last major conference: Some papers simply plant a flag — e.g., "popular method X applied to new task Y". In fact, it's not uncommon to see several papers with highly similar ideas at major conferences. In my opinion, it is not only difficult to keep up but also a net negative to try to, excessively. The problem at the core is feeling like you have to read all papers in the world.
Don't spend time trying to keep up with every paper. Don't rush research for the sake of flag-planting. Don't make decisions with the goal of being "in". Even if you don't buy into this mentality, there are plenty of non-PhD paths to research in the AI/ML field, as we discuss next.
❌ I need a PhD to conduct research in industry. There are famous researchers without PhDs, like Chris Olah. Generally speaking, these folks are the exception and not the rule, but it's possible to conduct research all the same without spending many more years in school.
- There are non-traditional tracks to an industry research role, via industry "residency" programs. These have been offered by Google, Meta, Apple, OpenAI, Nvidia and more (you can find live postings in a highly-watched awesome-ai-residency repository). Ultimately, this residency is a fast track to quality research. In turn, quality research stands on its own, with or without extra degrees.
- Machine learning engineer positions can also kickstart research careers — these are often opportunities to work on models or infrastructure. Although some positions require a master's degree, many others don't. At some companies (Google Brain, Deepmind, OpenAI), machine learning engineers and research scientists collaborate. At Meta, machine learning engineers can pioneer their own papers. At Apple, the difference is nominal.
- It's worth mentioning I disagree vehemently with gatekeeping scientist and machine learning roles behind a masters or PhD. I'd like to think hiring managers can recognize and will hire strong talent without minding pedigree. However, the unfortunate truth is that recruiters—who have to sift through hundreds of resumes daily—use these degrees to filter applicants for a first pass. In this regard, this is where Tesla Autopilot shines: Tesla will hunt for talent in all nooks and crannies, regardless of the degree and school. In fact, a PhD is detrimental as it suggests a candidate with more academic than industry mindset.
In sum, advanced degrees aren't really needed, but it does make landing the first interview easier in many cases.
❌ Engineering is boring and the only alternative is research. I used to believe this — that software engineering is boring. Put in more derogatory terms, software engineering seemed to reduce you to just a cog in a machine. It turns out, and as I learned over time, this isn't true; the research role comes with some more freedom built-in but a top notch engineer can just the same pioneer their own projects, pitch new problems, and spearhead innovation in some area. This is also why industry veterans are well equipped to do a PhD — they know the problem space, and they know how to make real impact on products today; the PhD then gives them a chance to apply their expertise to educating the academic community. Furthermore, you get more freedom in any capacity at a startup — researcher, engineer, designer, whoever you are. So it turns out that "boring" or "interesting" for me, was defined by freedom to pursue problems I felt were highly impactful. By this definition, engineering and researcher roles alike are both viable options. Granted, research roles are by definition a bit longer-term and less focused on immediate product needs, but neither role is inherently more "boring" or more "interesting" than the other. Both can live on the cutting edge.
Don't spend time trying to keep up with every paper. Don't rush research for the sake of flag-planting. Don't make decisions with the goal of being "in".
There are a long list of reasons to not do a PhD; funnily enough, ask any graduate student whether or not it's worth doing, and the answer is almost often a resounding no. I've to-date never heard a graduate student recommend following their footsteps. However, there are also very common — and very wrong — reasons for avoiding a PhD program; you'll often hear these as cons:
- ❌ I need to know my research focus before I start. Technically, the same problem applies to industry jobs, but industry roles on balance are better prepared to train a generalist engineer for a particular role. Knowing this, its actually true of both academia and industry, that knowing the problem domain you'll tackle is critical to long term growth. For your PhD program, one piece of advice I received is to go to industry for a few years first, to find a problem space that excites you. I still believe this is wise advice — it puts you in a stronger position to start and finish a successful PhD. However, it's also perfectly reasonable to start without knowing your ultimate focus. You could spend a few years in industry with a commitment to learning about various problems, then start the PhD; or, you could simply spend the entire time in the PhD fully focused on finding and then studying a problem domain. Neither approach is inherently worse. For me personally, I had a feeling if I went to industry first, I'd never a) want to go back to school or b) get a lucky admission a second time. As a result, the natural choice for me was to spend time in the PhD finding my passion.
- ❌ I don't belong, so the PhD is not for me. The irony about impostor syndrome is that everyone feels it at some point. So, if everyone thinks "I don't belong", then who does really? This applies to academics at all stages — applicants, PhD students, post docs, professors, and more! In every one of these categories, I've spoken to someone with a serious case of impostor syndrome. All this is to say that impostor syndrome is if anything a rite of passage — not a reason to dismiss the possibility of an academic career. This may appear to contradict with my last advice in the previous section, where I advised against doing the PhD to be "in". However, there's an important distinction between the two: Impostor syndrome should not stop you from pursuing the PhD, but it should also not be the motivation for doing one. In fact, impostor syndrome should not play any key role in your decision making.
- ❌ I'm making a financial sacrifice. This is true in many cases, and it's definitely true for the duration of the PhD. However, taking the time to become an expert can be worth it financially. I tried my best to estimate my own financial cost: I estimated my earnings trajectory had I joined industry right out of undergraduate. My total compensation now, after graduate school, is a solid 30%+ higher than my estimated compensation without a PhD. Furthermore, in about a year after graduation, my current trajectory with the PhD will surpass the industry-only alternate universe, financially, after accounting for the 5 years since finishing undergraduate. There are two factors that made this possible: (1) I was offered and accepted a full time startup engineer position concurrently, and (2) I happened to graduate and interview at a point when hiring was exploding "post-COVID" during spring 2022. This is not to say that you should bet on a PhD for financial reasons simply because I got lucky — I'm taking a much softer stance and arguing that the PhD doesn't guarantee financial doom.
- ❌ I'm sacrificing my social life. The PhD certainly isn't a 9 to 5, and the concept of a weekend doesn't fully transfer from industry to academia. However, I still had a ton of fun during my years in graduate school: I certainly made many many friends I would not have otherwise, and we hung out tons. With friends, I visited all the local hotspots near Berkeley — Yosemite, Muir Woods, Monterey Bay, Santa Cruz, San Francisco, and Lake Tahoe. I also traveled far and wide — Cabo, Hawaii (2x), Taiwan, Boston, Los Angeles, Austin, and more. I picked up plenty of hobbies — singing, taekwondo, content creation.. I also raced a duathlon! I'd say that's a decent amount of fun and socializing for 4 years.
Impostor syndrome should not stop you from pursuing the PhD, but it should also not be the motivation for doing one.
To be completely honest, I had no particularly strong reasons to do a PhD. I in fact started it believing I would not finish. In the end, I picked based on gut, and my gut said that the PhD would be the most new and exciting possibility. To help you in your decision, here are two pieces of advice that I realized in retrospect:
- Pick the option that opens the most doors: If I were to go back and re-justify my choice to myself, with the power of hindsight, I'd say I asked myself "which option opens the most doors?" At that time, the best option was to pick both the PhD and a concurrent startup offer. In particular, I was fortunate to be offered a full-time startup position that permitted me to take on the PhD concurrently. Furthermore, my manager—former numfocus president Andy Terrel—was a former academic. This was combined with a fairly industry-minded adviser—Professor Joseph Gonzalez—who himself spun out a startup right after his PhD. This lucky combination meant I could make the most of my time, learning about both industry and academia. In light of my objective to explore as many options as possible, taking on both offers was a no-brainer. I talk about my decision process in more detail, in How to make big decisions.
- Focus on one commitment (but I didn't). It's worth noting though that two commitments doesn't halve your output for each. It's much worse: You output maybe a quarter of the work for each. Across both full-time commitments, you're outputting half of the work you could have otherwise. The overhead of context switching and mentally juggling multiple responsibilities across both occupations is a lot to handle. As a result, funnily enough, if I were meet someone like my past self today, I would offer the same advice I had received at the time: Focus on one thing. Pick the PhD or the job and use the focus to triple your output. With one small caveat however: Since I kept multiple doors open, I found what I'm most passionate about, and that alone justifies the stress and confusion.
In sum, I had no particularly strong reason to pursue the PhD when I actually accepted the offer. In hindsight, I can certainly rationalize it and pass on the rationalization. As long as you didn't find any red flags above, I would say you're off to great start deliberating between PhD or not. To get a glimpse into how you'll grow as a PhD student, see What defines a "good" researcher? To get a preview of your post-PhD self, see What I learned in my PhD.
Pick the option that opens the most doors.
With that said, It's worth clarifying that "I just feel like it" is a perfectly good reason to pursue advanced studies. I don't claim to have had a reason myself, nor am I suggesting you need one! ↩
Want more tips? Drop your email, and I'll keep you in the loop.