The Precog Amplification

The summer break after graduation is when one realizes that IIIT has changed their life forever. It’s too soon to say whether it’s for the good or bad, but “sitting idle”, “not learning”, or “not chasing something new” become the biggest worries of life. Fear not, the bouts of peaceful wondering (and guilt-free procrastination) catch up soon, but for me it was the former set of feelings that saw me hunting for something to do in the summer.

I got a glimpse of what working at Precog would be like during PK’s DHCS course that I took in my final semester. It was meant to be a light course that undergraduate students crave for in their final lap, rightfully devoting the residual time appeasing their friends before the great dispersion. Ironically enough, it turned out to be anything but light, though still served the purpose of letting me have a great time with my friends. BBI (Building Better Interfaces), the conclusive project showcase, was almost like a start-up convention on steroids with students going to unfathomable lengths – pitching their projects and getting validation on their design process from practically everyone on campus. It’s difficult to forget a course such as this where you have witnessed banner wars, basketball challenges, beer pong and simply students going all-out on their projects. The enthusiasm was contagious and getting involved with Precog over the summer definitely seemed like an option to consider. Despite the fact that the domain of social {computing, networks, systems} was completely alien to me, I was incredibly lucky to make it through as a summer RA (maybe I just rode along on our cool DHCS course project, Fettle).

The following month, I started at Precog on a development project alongside a bunch of enthusiastic interns. During the first few weeks, as I familiarized myself with the domain of social media analytics, I found myself get attracted towards a particular thread. Given the rapid rate at which media content is growing on social media (~2000 images per second!), it was a question that often found its way into discussions  – how can we summarize this enormous dataset of social media images and make it more succinct and browsable? Having a background in Vision and an inclination towards research I found a certain affinity towards this problem and I shared my intent to work on this idea with Sonal, another RA at Precog who had just wrapped up one of her own projects (and happened to be looking for a new problem to work on. What luck, right?).

sometimes having the right answer is less important than seeing behind someone’s eyes why the question had to be asked – source

As both of us delved into conducting a high-level literature survey, we found that even though image data set summarization is a well-researched problem, in the context of social media data it is almost unexplored. I was all in for pitching the idea to PK right that moment, but Sonal, the more seasoned Precoger, advised against it and proposed for preparing a more polished case for the problem, one that would more eloquently bring out some exciting use cases. This was the first time I got introduced to the concept of making people excited about your research  and it starts with your PI itself. The first question PK would ask is “how do I sell this?”, which would encapsulate the other fundamental questions, “who are we helping?”, “do they need our help” and “how can we help?” (in that order). As brutal and business oriented the line of questioning would seem, I could appreciate the intent behind it. It was not meant as a discouragement of open ideas but instead as a first-round validation of how well the involved people are able to make a case that the idea is worth pursuing. This constant reinforcement that the job of a researcher entails being an effective communicator and convincing an audience that the problem is worth solving was something very unique to PK. Once we had this part out of the way over the many sync-up sessions, the ecosystem was made extremely conducive towards carrying out the required research work. Instead of narrating the experience any further, I’d rather break the rest of it down into more consumable nuggets –

1. The secret to all material success is self-discipline and grit – Be it your grades, getting an internship, an admit, building a project, coding a hack, or writing a research paper. If you can’t invest the required time and effort, it is wrong to expect a meaningful outcome. Yeah, you may get lucky once in awhile, but as Deadpool says “luck isn’t a superpower”. The 80 hour highly-organized work weeks at Precog, make sure that there is minimal dependence on luck. Keeping up with the expected commitment, Sonal and I continued working on our submission even after our RAship was over and saw it getting accepted to ACM Multimedia (you can read about how we approached our summarization problem and created #VisualHashtags here). This was one of the most rewarding experiences and only Sonal will remember burning all our stipend on Starbucks coffee, feeling guilty about constantly overloading their free wifi.
2. As for non-material success, it is empathy and gratitude – Academia is a very competitive domain and one where no matter how much you accomplish, self-doubt comes in plenty. Peer-review doesn’t stay limited to academic manuscripts and becomes a part of everyday life. It becomes important to support your colleagues in their effort because with so much competence around it is often that one starts getting really hard on themselves. Precog is one place where you would always find someone or the other to celebrate something as small as a midnight bugfix with. You would have to be seriously off the grid if you haven’t seen PK leading from the front, encouraging and taking special pride in bragging about his students and their work.
3. Diplomacy isn’t really a treasured asset in a research lab – The lab is devoid of any echo chambers because there are just so many strong voices. The senior most PhDs and fresh interns alike, everyone enjoys open channels of communication and get to navigate their journey at the lab. It was this environment that allowed me to switch projects with little friction and pursue a new idea during the middle of my tenure. It would be safe to say that with everyone here being absolutely blunt about their work and also with their feedback, I have started to adore conference peer reviewers (just kidding). The many reviews and rebuttals from Prateek, Niharika, Srishti, and Anupama greatly strengthened our ACMMM submission. (On a side note, I do think I owe an apology to Anupama for not being so server savvy at the time!)
4. Collaboration is equally important as individual brilliance – Working in the domain of Collaborative Cognition, I can vouch for the fact that collective intelligence trumps individual effort in more ways than just performance (Spoiler Alert: Avengers: Infinity War is an exception). An expert in one domain need not be an expert in another, and that is how it should be. At Precog I learned that it’s a big fallacy that one skill is better than the other. It may be more valued than the rest in a given context, but then it’s a matter of finding a match. What matters is being a master of that skill. The knowledge sharing that happens when different people, each with their own niche, work together leads to diverse perspectives and hence, exciting prospects. Such collaboration is common to most projects at Precog; even including our work on #VisualHashtags, where we had AVS and PK, two experts in their respective fields, collectively advising us on our research problem.
5. If something doesn’t make you anxious, is it really worth doing? – It is evident that people at Precog go places. Besides the qualification and merit gained at the lab, this success rate is because of the step-outside-the-comfort-zone attitude inculcated by PK. The bottom line is – a bunch of rejection emails in your inbox is much better than having only million dollar cheques from the Prince of Nigeria. Last year, I took the leap and applied for a few graduate programs in my area of interest – and to my delight received an admit to the Masters of Science in Robotics Program at CMU. As I start on this new endeavor this fall, PK being an alumnus at the same university makes it even more special – I am sure his mentorship and my association with Precog will continue in some form or the other.

Though my stint at the lab has been shorter than most, my blog entry and learning has been not. So, TL;DR: Precog is definitely a place to spend time at if you are remotely interested in rising from being mere nodes in a social system to being its philosophers and problem solvers. The line of research is highly interdisciplinary (Machine Learning, Natural Language Processing, Data Mining, HCI, Social Computing, Privacy and Security) and if you look around the projects here, you are bound to find something of your interest. It has everything to offer, from savvy GPUs to savvier researchers, and from cool projects to cooler friends. If you are ready to put in the hours, it’s an investment you won’t regret.