Why do a PhD?

In our second year of undergrad, my friend and I were sitting in a Thermo 1 tutorial being taught by a PhD student. At some point, we learn, that said PhD student has been at McGill a whole ten years. Ten years? Man. What a loser. We will be out of here in no time. We started in 2015.

In 2026, my supervisor reposts a Ben Recht blog on our Slack channel, decorticating the use of LLMs in peer review, and why the system is largely broken as is. The status of papers as CV success markers generates an overwhelming tide of submissions and reviews that are, in some cases, as good as random.

What is then the point of peer review? Moreover, what is the point of doing science? In particular, why do we do a PhD and learn to navigate this system? What do we really get out of it? I put some of my thoughts below. Of course, they only reflect my own experience having done a masters and (almost) a PhD in robot perception and are not meant as an absolute truth.

Let’s address the elephant in the room first: Money. My understanding from discussions with colleagues is that, purely from a financial standpoint, the salary gains from a PhD do not outweigh the four years of masters-in-robotics salary that we miss out on. This excludes some narrow fields like autonomous vehicles where a PhD can actually contribute to a very well-paying position. I think this is even more true for the general technical field, where masters plus four years of experience gets you financially more ahead than a PhD.

Second, why did I in particular do a PhD? I was on the fence entering my second year of masters. I did not really know what I wanted to do. Then, on the NSERC PGSD application deadline I get a slack from my supervisor - “The deadline for NSERC application is TODAY”. And I thought, what the hell, I’ll apply and see where it gets me. The rest is history. So, for me, the honest answer is: academic inertia. I believe I made the entirely correct decision, but not really for the right reasons. There are certainly people who crawl out of the womb doing hackathons and making raspberry-pi roomba variations to terrorize the household cat. I was never one of these people. I like math and I like applying myself.

So, what is a PhD? At McGill, we have four deliverables: a literature review, a thesis proposal, a preliminary exam, and a defense. The interesting bit is that these deliverables are almost purely checkboxes. They are only intended to measure a) whether you are capable of doing research and b) whether you have done research/something novel. The day-to-day life is you sit in the office, and you try to figure out the answer to the famous six questions

1. What is the problem
2. Why do you care
3. What is the state of the art
4. What is the problem with the state of the art
5. What solution do you propose
6. What evidence do you have that your solution works.

As a student who knows next to nothing, this is hard. You do get some guidance. But a common misconception is that someone will tell you what to do. The truth is that nobody will. This is not a bug, but a feature. It is up to you to pick a direction. This is one of the core differences with a typical masters’, in that during a masters there is usually more structure and direct supervision. In a PhD, you answer these questions for yourself. You have to tell yourself what to do. And truthfully, at the beginning, we all suck at it. However, how do we get better? How do we improve at being able to say “I want to do this, and this is why I will do it”, when investigating a new subject that we do not understand? With practice. Repeated practice. We try one problem. We fumble, we do not read the literature well enough, or we do not push the research as far as we should. That’s fine. We have time. We do it again. And again. And again.

I believe this is what truly differentiates a PhD graduate. It is not just technical know-how about their specific field, but the practiced ability to figure out what next step to take in an unknown field. More colloquially, to know what they want, purely by virtue of practice and showing up to their supervisor meeting having to say “This is what I want to do”, and being able to defend it.

On education versus training. What is training? Training is the process by which we become better at a specific task, such as coding or engineering.

What is education? Education is attaintment of knowledge and understanding about the world. Unlike training, it is not by its nature meant to be directly “useful”. In some sense, it is an addition of knowledge that shapes one’s worldview about the world.

I am reminded of another conversation with colleagues from many years ago, where they mentioned the horsehoe theory of doing a PhD. Whether a PhD is done in arts, science, or engineering, the core skills that it develops tend to being the same. The same ability to sit with a subject for a long time and question it, leaves the student with a similar spirit regardless of the specific domain.

I remember the following quote, which I almost certainly propose out of the original intended context, and is by a famous Russian linguist Dmitry Likhachev. “The fundamental principle of intellegentsia is intellectual freedom, freedom as a moral category. An person of intellegentsia is not free only from their conscience and their thoughts.” -Dmitry Likhachev

I did not use any LLMs to write the post itself.