If there is one thing that I have taken away from my 3 years studying CS at college, it’s that learning is not exclusively reserved for the lecture halls. In fact, I would argue that I have learned more by myself in my bedroom in the past 10 months than I have in the first 2 years of university.
There are several factors that went into this, but I want to focus on my experience doing research for both the computational brain lab and machine learning (ML) lab at my college. Before that, here’s a little background.
I breezed past the typical undergraduate CS courses and to be honest, I can’t say that I learned much more than I had for free online or that I could have in far less time for free online (do you see the pattern here?).
Learning what IEEE-754 floating point representation is was great and all, but something else caught my eye the winter of my sophomore year, ML. I started going through Andew Ng’s famous CS229 on coursera and was introduced to a whole new world with these interesting concepts. I quickly became obsessed. But at the same time, I stumbled upon the world of entrepreneurship and listened to countless founder stories of successful startups. My university classes became secondary to what I was teaching myself online and what I was consuming via podcasts and founder interviews on YouTube.
Now for those of you who are being swayed to skip out on university because of what I’ve said until now, please don’t take that as the main message. I am blessed to be able to attend a university and get this experience, and I will forever be grateful and remember these years of my life. What I do want to say though is that you will find much more fulfilment by proactively challenging yourself to find things which will satisfy your intellectual and/or creative desires beyond the lecture hall.
With my new found interest in ML, I scoured my university’s computer science website for things that an undergraduate student could do for more exposure to AI and ML. I came across the term research, which was obviously a term that I was familiar with but I did not know that the world of AI research was so vast, with its history going back as far as 1955. There were so many professors who were working on incredible things, so I hastily read a few of each of their publications and cold emailed close to 10 professors, each with a genuine interest in their work and desire to join their lab. 9 of them ghosted me, but 1 professor emailed me back saying that he wanted to meet the following day.
I had finally found an outlet where I could work on some more interesting things related to AI. Turns out, I had gotten far too ahead of myself. I was assigned to work on an interface for a robot head, which I happily worked on but quickly found to not be the kind of work I was looking for. Then covid-19 hit like a truck and all students and professors were sent home, and here we are now.
Covid didn’t stop my research group, and of course classes were still in session. I apologize for potentially sounding insensitive, but to be honest Covid brought a turning point in terms of research for me. I constantly started bugging a PhD student in my lab about what I could learn and work on during this time working remotely. After several back and forth emails, he offered a chance for me to work on an actual AI project to work towards a second author publication with him and one other lab mate.
Working on this publication quickly took all time away from me, and I loved it at first. The intellectual challenges that I faced in having to learn so many things on the fly fundamentally changed the way that I learn and I will forever be grateful that I learned how to learn during my time working on this project. Working with two other PhD students, we successfully submitted our project to a conference called CoRL and the work was accepted. Great! But I faced several realizations about the world of AI research, at least in academia.
I realized that research in AI was really cool and intellectually challenging, but I also realized that it took a very large amount of time away from both me and the PhD students in the lab. Deep down inside, I knew that I always wanted to start my own company, and I missed listening to and studying how other founders started their startups. I was passionate about building, but research had taken the time that I once spent listening to CEO interviews and podcasts away, not to mention time I once spent learning how to build software products. But the prestige and challenging nature of research and PhD students was compelling to me and it seemed like an easy route to respect from friends and family.
Amidst this internal conflict, I reached out to the ML lab at my college because it would be great to have another publication and potential recommendation letter for applying to PhD programs later on. I joined the ML lab November 2020 and worked on some very interesting things in probabilistic generative models and representation learning. The work that the lab was interested in was exactly what I had wanted to work on when I first fell in love with ML and AI. However, every day that went by reading a few papers a day, preparing for presentations on some of these papers, and going deeper still into the rabbit hole of ML research, my desire for making products and working on being an entrepreneur screamed louder and louder.
I took time to carefully reflect on what my desires were, and I found that all I desired was to live a life of continual learning and working on interesting problems. Research was a means by which I was gaining this for sure, but my opinion on research started to become exactly like what my opinion on classical lecture learning was; it was slow and I found that I could do things faster on my own. Lightbulb I realized that the bureaucracy of academic learning and research put ceilings over the potential that I believe I have. Only a month and a half later I contacted the professor of the ML lab that I wanted to work on my entrepreneurial dreams and that I could not commit to contributing to the lab any longer.
And here I am now, writing this blog to share my experiences with the people of the internet. Some of you may be wondering, do I regret giving up my chances of going to a prestigious PhD program and gaining the “elite” status of being an AI researcher at a top university lab? To be honest, I don’t have an answer for that explicitly, but I do know that the past month where I’ve been preparing to launch a startup with my cofounder and closest friend has been just as challenging and just as fulfilling, if not more, as my experience working with the computational brain lab and ML lab. The ceiling I once felt from university and research is nowhere to be seen, because the path to entrepreneurship is impossible to foresee. Everyone’s story of launching a successful startup is different, thus I live everyday in mystery just trying to get closer to my goal of working on challenging problems to help people live easier and happier lives. The path to being a researcher is quite established and oftentimes a rinse and repeat cycle. You do research in your undergraduate years and make strong relationships with your professors in hopes of getting strong recommendation letters for PhD programs, then you hopefully join a top 10 lab in the world and do research for 5-6 years before either doing a postdoc at a university or joining an industry lab. Then you work for the next few decades until you become a tenured professor at a nice university while still doing research and guiding younger researchers to push out publications for your lab. This sounds great for sure, but living in uncertainty sounds even greater to me. Even if I don’t gain a single ounce of respect or prestige that I could have from continuing to be a researcher, I will have no regrets because in the time that I live I know that I will be working on creating products and services to make the lives of people just like me easier. There is no ceiling to entrepreneurship. You control what your path is, and I’ve just started going along my path after taking a few detours.
I am by no means telling people to quit their dreams of pursuing research in AI, rather I am sharing the reason why I both joined and quit research in the field as an undergraduate student.