Ishan Kapnadak – Indiana University

Introduction

Hello everyone! I am Ishan, a third-year student of Electrical Engineering, pursuing a (double) minor in Artificial Intelligence and Computer Science. I’m also a music buff and occasionally make music. For the past few months, I have been interning in the field of Cognitively-inspired Artificial Intelligence at Indiana University, Bloomington. This is my journey.

Background

Research has always fascinated me from the time I set foot in IIT. Exploring uncharted territories, discovering new stuff, and breaking through boundaries was always something I looked forward to. So, when the intern season came along, a research internship was a no-brainer. But a few weeks in, I began to have doubts. With all my peers around me bagging high-paying corporate internships and preparing for tests, interviews, and coding, it occurred to me that perhaps my desire for a research internship was a bit impractical and naive. 

I ended up applying to a few companies out of peer pressure. I was rejected by all of them. Hope turned to despair and then to frustration. I had almost given up and convinced myself that I probably wouldn’t get an internship for the summer. I made a few half-hearted attempts at the dreaded apping process. The fact that I did not hear back from any of the professors only strengthened the belief that maybe I’ll end up without an internship.

Now was the time where I would blindly scroll through the PT-Cell website, looking for anything that might remotely interest me. By pure chance, I happened to stumble across an offer from Indiana University regarding a research internship. We were required to select a professor (or two) of our liking, and submit a statement of purpose regarding our motivation for the internship. I scrolled through the list of faculty and found one that I liked. Although it seemed like an advanced area of research, I thought that I should give it a shot anyway. I wrote up my statement of purpose, signed the IAF, and forgot about it. 

Fast-forward to a couple of months later. I receive an email out of the blue informing me that I have been selected for the internship. It took me a few minutes to process what it meant. There was no second round where I would falter and lose the internship. This was it. I got it. Once the realisation hit, I was elated. When I told my parents, they were just as elated. In fact, my mom started calculating the cost of living in Indiana and was concerned about the weather there. However, when I read the acceptance letter more carefully, I noticed that the internship was virtual! So all my fantasies of travelling halfway across the world came to a standstill. Nevertheless, I was excited.

The First Meet

I have been working under Prof. Zoran Tiganj, Indiana University in the field of Cognitively-inspired Artificial Intelligence. We had our first Zoom meeting around the 14th of May. It was largely informal and more of a getting-to-know-each-other session. The professor was really friendly and seemed just as enthusiastic about the internship as I was! We discussed the Covid situation in our respective countries and talked about our experiences in a virtual classroom. We then talked about my expectations from the internship, and if there was any topic that particularly fascinated me. Since the field was new to me, I had absolutely no idea about the kind of research that goes on in the field. 

To give me some idea, the professor sent me a huge list of 65+ recent research papers in the field. The aim was to read the abstracts of all these papers and zero in on a subtopic that I found really exciting. The first thing I noticed was that all these papers were really recent, with some being published as recently as April 2021 (just a month before the internship began!) I soon realised that this was a really up-and-coming field. Personally, this was exciting because I felt that I had an open canvas to paint on – that I can contribute significantly to the field since there is so much to explore! Although I found a lot of the papers interesting, I decided to focus on the problem of Vector-based Spatial Navigation. 

My Project

(Feel free to skip this section if you don’t want to go into the technical details)

Mammals, particularly humans, are able to flexibly navigate the world they live in. We are able to explore new areas, return to remembered places, and find shortcuts from one place to the other. Although these abilities feel natural to us, it is startling how complex the underlying process is and how easy it seems for our brain to execute this. 

For artificial agents, however, spatial navigation is a challenging task where they are absolutely no match for humans. That’s because humans have these magicians -specialised neurons that make navigation intuitive and effortless – in their brains. Two kinds are of particular importance – Grid cells that fire at regular intervals as the mammal traverses through space and Place cells that fire at some very specific locations. Tons and tons of these kinds of cells are what allows our brain to calculate distance and direction to the desired location, and effectively allow us to navigate. 

A recent DeepMind paper sought to test the theory that grid cells support vector-based navigation, wherein they developed an artificial agent which learns to localise itself in an environment. They found that grid-like representations spontaneously emerged in the network, and were crucial for navigation. This problem really fascinated me. In the first few weeks of my internship, I implemented the DeepMind paper as it is. And indeed, grid cell patterns emerged spontaneously! My next step was to play around with the network and observe how the results change. I ended up getting some nice results that were consistent with a few computational models about this problem. Towards the end of the internship, I tried to solve the same problem using a different method. My professor and I felt that the network in the DeepMind paper was too ‘constrained’ to solve the problem in a particular way. We were interested in observing if the grid cell patterns emerge spontaneously even with lesser constraints. 

Since the project was getting really interesting, my professor and I agreed to extend the project indefinitely!

My Experience with a Virtual Internship

Prof. Tiganj and I held meetings roughly weekly. Although time zones were a challenge, we settled on a time that was convenient to both of us. The general format of the meets was largely constant throughout the two months. I would update the prof on the work I had done in the previous week, and the results I had observed. We would then discuss new things I could try. Inevitably, each meeting also had several digressions where the professor would discuss some of his current research, and give me some psychology and neuroscience tidbits. This made the meetings really fun and enriching. 

Though an in-person internship would have been my preference, a virtual internship had some of its own advantages. First, I had a lot of flexibility in planning my work. Since the internship was virtual, I worked on my own time and sent some regular updates to the professor. This fit in very well with some of the other activities I was pursuing in the summer. 

And second, it gave me a lot of independence in my work. This ultimately helped me to face challenges on my own rather than taking help from the professor or one of his students (as I might have done in an offline internship). 

On the whole, while an offline internship would have been a newer and more exciting experience (not least of which was a chance to stay and work in the United States), the online format has been a super fun experience for me.

My Perspective on Research

Before I started my internship, I had some preconceived notions about what research would be like. I used to think that I would spend hours and hours brainstorming, and change the face of humanity with my work. Of course, that is not something I did. Research seemed more like hours and hours doing mundane work, with occasional spikes of excitement and ‘Eureka’ moments. But those moments more than make up for the generally frustrating period of experimentation. After a few months of research experience, I can still say that research is something I am passionate about and want to pursue. 

Takeaways

Back when I was in school, I hated biology. Dreaded it. Despised it even. But somehow, I got through it. I remember my grandmother (a doctor herself) telling me countless times that no knowledge is ever wasted. I never agreed. Why would I need to know how the brain works? And yet, four years on, I found myself investigating exactly how the brain works. Indeed, no knowledge is ever wasted. You just don’t know when you will need it. 

I also learned that faith and belief are crucial. Whether it is an internship, or work, or just life. Challenges will come and go. Failures are inevitable. But something better is always there in store. Had I been accepted into any of those corporate internships, I wouldn’t have gotten the opportunity of doing something I really love. At that point, it seemed like a failure. But now I have learned to accept that there is no failure. Only delayed success. And success is whatever you define it to be. 

I realise that my story is not your conventional success story. But it is definitely my success story. The one thing I have learned through this whole experience is that your journey is your own journey. Unlike what Robert Frost said, there are not two, but countless roads that diverge into the woods. And it is up to you to choose which road is yours. The one less travelled by, or the one everyone takes. Because it is yours to choose. And the one you choose will always be the right one for you.

3 weeks ago

Leave a Reply

Your email address will not be published. Required fields are marked *