Me and Research
I started doing research in the summer before my senior year of high school, the summer of 2019. I’m closing in on 4 years now. The time I have been doing research subsumes the COVID-19 pandemic as well as my entire undergraduate education at Georgetown. I was 16 when I started doing research and I am 20 now. If I have an expected lifespan of ~80 years, then I hope I get to have 60 more years of research left in me.
To be completely honest though, I got into research because I wanted to maximise my chances of getting into MIT. Those days in high school were not so long ago, when dreams had a timeline extending to college and no further, when everything seemed so immediate and important. Fun fact: I did not get into MIT. Thankfully, research has stuck with me though.
It infuriates me now how cynical my original motivations were. My attitude towards major life decisions as a high school student bordered on careerist. I wanted to do things for the sole reason that they would look good on my application. I had very little respect for what I learned in school beyond the final grade or AP score, even though I did pretty well. I also had an inflated sense of my own abilities—it got better after I moved to D.C., a bigger city where people had greater aspirations and the impressive accomplishments to back them up. But I really did expect to get into any college I wanted.
I don’t think such an attitude is uncommon among many high school student in the United States today. The nature of college admissions has a big role to play in this; people optimise for what is rewarded, and when there are a lot of people and very few spots the optimisation gradient blows up. It does kind of suck though.
What even is research? I don’t think I have a great answer. I think the name for it carries a lot of baggage. Maybe at its core, research is just about asking questions. Just a fancy name for applying one’s innate sense of curiosity. Unlike much of what we deal with in everyday life, the search space in research is infinite. Everything I did in life before research had a finite search space, in contrast—there are $n$ prestigious colleges that are worth going to, there are $x$ acceptable fields of study (and they are all STEM), there are $y$ places worth working at and the metric to optimise is a 1-dimensional scalar called “salary”. Such a big world with so little that we understand, and we carve out tiny domains to fit our whole lives into.
At the urging of my mom,1 I cold-emailed a lot of people in the summer of 2019 to try and get to work on a research project. I had some vague idea that I wanted to do computer science stuff, but also with linguistics if possible. I emailed a professor working on compilers, some lab in D.C., maybe a couple others I don’t remember. Funnily enough, only my first email got a response, the one I was most hoping to hear from— Nathan Schneider at Georgetown University, who worked (and still works) on computational linguistics. About a week passed between the email and the response, and then I responded immediately on July 2nd. I didn’t hear back for 3 weeks, and got pretty worried because I was starting to get excited about it! Fortunately it worked out, and I got to meet Dr. Schneider and his Ph.D. student Luke Gessler on August 5th, 2020.
My first research project was on computationally modelling schwa deletion in Hindi, i.e. throwing machine learning methods at the problem and hoping it works.2 This was my first time doing a real machine learning project. I had played around with neural networks once before, for my final project for the online class CS50x where I first learned programming. It was a really dumb neural network for playing Connect 4, and since I didn’t know what backprop was I implemented “genetic algorithms”, a fancy way of saying I tried a bunch of random initialisations for the network weights and had them play each other. It was really fun. But I digress; now, I got to apply neural networks to a real problem.
The project was pretty successful if you look at the metrics. We beat out a lot of previous hand-crafted methods in the literature and I got to write my first paper—well, it was more like “learned to write a paper from the example of Luke and Nathan and wrote a few things here and there myself”—which ended up getting accepted at ACL. But I would not say I fell in love with research at that point. I liked the work and I definitely preferred it to schoolwork. I guess I would say that I liked working on hard problems, and this was a hard problem. Solving it was satisfying.
The first research project I did where I truly felt driven by my own curiousity about it was extending SNACS to Hindi. I spent the summer of 2020 working on SNACS annotation of the English version of The Little Prince with some of the grad students at Georgetown (I had gotten accepted there and committed for undergrad). I got the idea in May to try and extend this to Hindi, mostly out of curiosity as to whether it would really work, since Hindi’s case markers were way more complex than English prepositions. I wrote up a basic annotation guidelines document, did a bunch of example sentence annotations myself. Research is basically self-nerd-sniping and I had nerd-sniped myself for the first time. In fact, I did not realise that this is what happened until I sat down to write this. Unlike most things I had done before then, I really just was curious about this and had no ulterior motives driving my desire to work on the problem.
From that point onwards, everything that I have worked on has been motivated by my own interest and love for the problem. Sure, there are frustrating moments—sometimes there’s a seemingly unsolvable bug in the code, sometimes revising papers and dealing with collaborators is a pain, sometimes the problem doesn’t feel exciting. But, like any loving and healthy relationship, if it is worth doing you slog through the tough parts and are rewarded for it. It seems to me that the motivations for working on research should be a lot of curiosity (“Why?”), a belief in the importance of the problem, and a little bit of dissatisfaction with the status quo.
I basically just started research and don’t consider myself particularly good at it. I am still developing my research taste. I don’t think I’ve worked on particularly exciting problems thus far. However, I find this fact very exciting because it means the ceiling is so high I can’t see it, and I have room to grow and much cooler things left to work on. But even beginner researchers get to see amazing things—if you spend even a little time with information theory you get to grasp a little of the beauty of math, if you read the transformers paper and go through the implementation yourself the insights behind it leave you in awe. Especially working on problems in AI or adjacent areas right now, you get a glimpse of the future and it feels like there are too many exciting problems to choose from right now.
The rest of my thoughts on research were already better articulated by Richard Hamming.
I think it is very definitely worth the struggle to try and do first-class work because the truth is, the value is in the struggle more than it is in the result. The struggle to make something of yourself seems to be worthwhile in itself. The success and fame are sort of dividends, in my opinion.