Sunday January 10th
- I spent last week at WICSS, the Winter Institute in Computational Social Science, which was held at the University of Arizona from January 3-9, 2021. Okay, I lied. It was actually held virtually this year, although it was supposed to be at the University of Arizona.
- In a strange twist, I think I was the only computer science PhD person there, but we had a fair share of persons in poli sci, sociology, linguistics, and other such interesting fields. And I got a lot of insight from the people and topics covered.
- You can scroll down the list and see a photo of me here. I didn’t know that it would be non-CS people, so when asked for a bio, I just said my specialization instead of my major, which is funny. Oops. Oh well, live and learn, I guess.
Why was I there?
- I think we can’t just make technology without thinking about the people that are impacted by such technology, and so systems should be built bearing in mind the people (especially with domain expertise) who use them.
- So surprisingly, I applied, and got in (as one of 30 persons chosen). To get in I had to submit my CV, papers I had currently worked on (sample of published work, etc) and what I’m currently working on (proposal). It’s interesting because depending on your field, you probably might publish more in journals versus in conferences or conference workshops. But they were pretty flexible either way.
- Interestingly, the person I did a conlang class with completed his PhD in Linguistics (he works currently as a software engineer) at the same school, University of Arizona. Interesting. The person who taught our nlp course apparently is also a Haskeller, and even briefly spoke about Type Theory, which is funny.
- It also exposed me to some of my own biases with respect to access to information. I think in one of the socials with two other persons I was casually talking with, I found myself asking them why they didn’t contact the company? In CS, this is something that we do and people listen to us, because well, everyone wants to talk to CS people, so a lot of the time companies come to us, or even contact us saying that they heard we are currently working on a thing that they may be interested in. I realized that this wasn’t as straight-forward for every major. It can probably create this bubble where people in CS (particularly in ML or data science) can think that the work they do is more important, but companies wanting to pour money and give grants is a poor metric for this, based on the baseline that especially tech companies are already more interested in sponsoring and giving grants to CS research. Weird. It reminded me of a Maths PhD event I attended two years ago, where a friend of mine, said “in Maths, we are lucky to get a free pencil!”. It made me realize the disparity of interest and funding between the fields, which really incentivized me to think in the future of how we could increase cross-collaboration to work on impactful problems across fields.
So what was covered?
- We had insightful speakers each day. I loved the sessions on web scraping, surveys, nlp, machine learning, and gis. I guess that’s almost all the sessions haha.
- We had workshops, which were done with R code. Those sessions were recorded and there is a github repo that has the appropriate exercises, and as workshop attendees, we had access to the recording.
- I had a lot of fun in the “Wonder” rooms, which were these spatial chat rooms in which we were able to float around and chat with various people. I also got really great advice about where to find data for a project I’m working on. Most of all, the group seemed non-judgmental. I’ve experienced similar sentiments in the Maths and Linguistics graduate communities. It seems to be a part of their culture (well, the tiny bit that I’ve seen), as well. Once again, maybe part of this I’ve seen in CS are a result of the influence of, and dependency on, corporate funding. It’s a metric by which we can begin to “look down on” other labs that don’t have, etc. So the focus is taken away from “what are you working on, and why is that important?”.
- I haven’t used R in years, but it was wasn’t bad to get back into. I like that the community really is a stats-one. It’s really a language for statisticians but I had no idea that since I had picked it up, people were using it to nlp and all sorts of stuff like that. I remember that one guy in LA who even made a logo in R with R code and printed it onto a tshirt lol. I also still subscribe (and have been for years) R-bloggers, which is really great. It’s a thing I’ve casually kept up with reading whenever it is in my inbox, even though I’ve been writing a lot more Python when it comes to stats and ML.
Games and Socials
- Besides the Wonder room, we also had trivia games, and feedback forms to fill out each day. And they really did pay attention to the feedback. It was a really well-organized workshop.
If you’re thinking you might benefit
- Definitely apply! I got a lot of insight from the different perspectives of persons doing PhDs in non-CS disciplines. The purpose or motivation, even though there is still code and data analysis involved, is different. That was refreshing. There is less focus on the tooling and more on insights on data, which was (as I said) different and refreshing.
I have to head to sleep
- This week is treacherous! I am giving two talks, and one of the workshops starts at 3am, and another at 10am, so you can tell that’s kind of awesome. You can guess that somewhere mid-week I might crash and end up taking a mini-nap. I’ve had moments before where I had to turn off my video in a session and take a nap. I’m sorry…being human is hard haha. Video conferencing really really wears me out.
- This Holiday season I really enjoyed long days of silence and the time to do focused work. I value that the most.
- A bunch of books came in today; including puzzle books, ones on Privacy and Cryptography. I’m beginning to think I’m more of an ML person who thinks like a cryptographer, which makes sense, because my advisor’s background (and our lab) is primarily crypto. I think I like that a lot than being an ML-heavy person. I have real issues with say, a solely data-centric view of linguistics, and I’m not as interested in that as say, linguistics, in the same way I’d say I care more about how a cryptographer thinks than being an ML person first trying to apply cryptography. But that’s a discussion for another day. A lot of these fields have incredibly rich backgrounds that are incredibly interesting, even though they do need more people who really understand the domain and can apply ML techniques. I personally think that’s the better way to go, rather than the person who just gets off on ML and tries to apply it to everything under the sun. But that’s just me.
- (Coincidentally, this isn’t just me. I attended BUDS at NeurIPS 2019, which is a series of mentorship roundtables, and Dr. Bouman expressed similar sentiments. That is, that it is worth staying in one space long enough to understand it. Perhaps the people who do these kinds of breakthroughs have done precisely that; they’ve stayed with a problem long enough to chip away at it, understanding “the domain under the ML”, if we can put it that way.
- One of the things that is fundamentally interesting about (crypto / secure comp) + ML (or rather, that intersection) is that the former is about precision, and the latter is about approximation. So they seem almost incompatible. Anyways, that’s all I have to say on that.
And that’s it.
Written on January 10, 2021