aqua-ified thoughts

never developed nor fully solidified

Fall 2020 Reflections

Wednesday, December 30, 2020, 12:50 AM

Last time, I wrote about my first-month impression of the COVID-infested Fall semester. It has been three months since then, and I’m honestly surprised at how fast it was. Yet, simultaneously, each week felt painfully slow. I guess this is what happens when there are a lot of deadlines and your brain can only handle so many at once.

Enough rambling! It’s finally the end of the semester (and more importantly, the end of 2020) which means I get to write a real end-of-semester review. Because I talked so much about my thoughts on classes last post, I don’t really have much more to say except the things that changed (for better or worse). There have been a lot of developments during the semester, though, including some of the more personal ones. I’ve also collected some juicy statistics. So, that’s what we’re going to talk about.

How to sleep like a normal human

Remember the sleep log graph I posted last time? If you don’t (or haven’t seen it), here’s a summary: I slept during daytime. At first, it was something like 9 AM to 5 PM, then it kept shifting forward. The duration also varied drastically between as low as 3 hours a day and 13 hours a day.

Part of the issue was timezone misalignment: I had classes to take between 9 PM and 2 AM some nights. However, the lack of discipline also played a role. My body clock being severely misaligned with sunlight also made it worse. And quite frankly, when you have a schedule like that, you can’t even decide on when you want to sleep, because none of the choices make any sense.

Every once in a while I would have the motivation to fix it, and so I did. I was able to sleep at 1 AM like some other people for a while. And then it would be 4 AM. Before I knew it, I was back to sleeping at 3 PM, sometimes even skipping sleep. This cycle happened a few times.

If you know me personally, you were most definitely concerned and probably told me many times to go to sleep. If you don’t know me personally and read the last blog, you were probably also concerned. If you don’t know me personally and just learned about this, you’re now probably concerned too.

Today I have good news. Look at this:

Top: When I slept. Bottom: Total duration of sleep ending on that day.
(Credits to @arunwpm for providing me the visualization!)

Miraculously, I fixed my sleep schedule and managed to maintain it, not just for a week, but for a whole month. For the first time ever, when people ask me when I usually sleep, I have a definite answer. It’s 3:30 AM till noon. (Note that’s 8 hours!)

Here’s my theory on why I managed to fix it. From the graph, my sleep schedule was decaying extremely quickly in early November. It got so bad that I basically circled back to the “normal” sleeping times. That coincided with when I was motivated to fix my sleep schedule. So, it was natural for me to maintain it before I would sleep at, say, 8 AM again. I was still failing, though.

Then, November 20th, thanksgiving break started. Coincidentally, that was one week before Thailand Olympiad in Informatics, which meant there would be a training camp at a nearby university (about two hours away). As hosed as I was, I decided to commit to tutoring, hoping it would help distract me from all the school work I would have to deal with after the break.

That turned out to be a fantastic decision. I was forced to wake up consistently before 8 AM to get my breakfast (a meal I’d never had in a really long while) and lead my students from wherever we were sleeping to the university. That was a solid 20-minute walk, plus another 20 minutes back. Factor in my activity during the day then that week was probably the most active I’ve ever been in the past eight months. (Imagine having an average of only a few hundred steps logged on your phone every day, and then suddenly getting almost 10k steps/day for a week.) Add the responsibility to wake up on time every day, and there you go, a recipe for fixing my messed up sleep schedule and not being sedentary.

The camp was rather reinvigorating, so I had the motivation to somewhat maintain the schedule as I came back home. I’ve been getting loads of encouragement, and having someone hold me accountable works amazingly well. (Thank you so much! <3) I can’t say with absolute certainty this will last forever, but since I already got into the rhythm of forcing myself to sleep and wake up at the right time, I’m pretty optimistic this will keep going.

Being hosed 101

hosed (adj.) - flooded with work, as if attempting to drink from a firehose1

At the beginning, I was curious how time-consuming MIT would actually be. So, I decided to set up Notion2 to track my time. After a full semester of collecting data, I got a pretty revealing breakdown of how intense each class was. Take a guess at how bad things were. Recall, I took:

  • 3.091 Introduction to Solid-State Chemistry (GIR3)
  • 6.036 Introducton to Machine Learning
  • 6.854 Advanced Algorithms
  • 21M.011 Introduction to Western Music (CI-H4)

Here’s the big reveal:

Time spent on academics in Fall 2020, broken down by classes. Note that dropped classes include 6.046 Design and Analysis of Algorithms (did the first two psets) and a bunch of other discovery classes I tried for the first three weeks.

As it turns out, 6.854 actually took the majority of my time, which was kinda expected. I recorded a total of 256 hours5 which is just a little less than 20 hours per week. I don’t think it tells the full story of how intense the class actually is, though. I couldn’t predict the difficulty of the psets, so the added anxiety made the class feel like it actually took 30 hours per week. In hindsight, I should’ve spent less time worrying and more actually working on the psets. Also, lesson learned: asking for help does significantly cut down the amount of time. I wouldn’t have survived Pset 9 if not for kind folks who were willing to explain the same thing over and over to me. I’m forever grateful for your help. 🙏

6.854 also has a final project where you have to do one of these:

  • Pick a few research papers you find interesting and summarize and synthesize them. The goal is to present the concepts with most clarity and perhaps present new connections between those papers. The papers should be relatively new (i.e. published in the last 3-4 years).
  • Pick a paper (or more) containing new algorithms. Implement them to try and figure out their performances on actual machines by designing some test sets. Perhaps find optimizations that help the algorithms work faster in practice.
  • Do some original research.

From the very beginning, I planned to do the reading project. Even Karger himself recognized it as the most straightforward way to get an A. There’s one problem: The project was released super late into the semester, with proposal due right after Thanksgiving break and the project itself due the week after. We only got the info and the old, sample projects a few days before the Drop date (for dropping classes or changing grading system). The level of expectation came as a big surprise, and the time alloted for it was barely enough. Needless to say, the last few weeks of the semester were stressful.

Surprise! There was also a mandatory peer review/editing session two days before the project due date. I freaked out. Luckily, it only added about 1.5 hours, but this was the tipping point. I wanted to recommend 6.854 to people who love algorithms but now I realize there needs to be an enormous warning tag. Karger is an amazing professor who gives some of the best lectures I’ve had. He just needs serious help organizing stuff6.

So, this is what ended up happening: We looked for papers that seem promising for a few hours, chose them, and sent in project proposal on Monday (Dec 5). After two days, one follow-up email, and five other assignments I did while waiting, we finally got an approval. We only got to start working on the project on Saturday because, well, it’s the final stretch; things get busy (understatement). The project was grueling because the paper we chose was rather badly written. Slogging through7, we managed to finish it on Wednesday night, right on the due date8 (Dec 9). I don’t know how to explain the relief.

Here’s the breakdown of time I spent on classes each week as we’re moving away from 6.854:

Time spent on academics in Fall 2020, broken down by week. Note that the first week (with no psets) count as week 0. Thanksgiving and final weeks are merged into adjacent weeks. For smaller assignments, I defaulted to recording on the due week. For larger assignments, I tried to be as accurate as I could be.

Surprisingly9, 21M.011 took substantially more time than 6.036. I spent a total of ~80 hours, which is about 6 hours per week. On normal weeks, it’s actually about 3-4 hours. That’s 3 hours of a mandatory asynchronous lecture and two live recitations. Add another hour if you have a quiz that week and need to prepare for it10. The rest of the hours were thanks to the essays. 21M.011 is an amazing class with truly fantastic staff. (Emily Pollock and Teresa Neff are the best!) The workload is perhaps among the lightest of all HASS classes11, so please consider 21M.011 if you’re still looking for a fun CI-H12!

6.036 had 60 hours recorded, or about 4.5 hours per week. 6.036 was perhaps the most consistent class in my schedule, mainly because there were no midterms or final exams and homework take roughly two hours every time. Add two hours of weekly labs plus exercises plus nanoquizzes. Then maybe add an hour or five if the god of bugs are in a bad mood or you want to go down the matrix calculus rabbit hole.

Finally, we get to 3.091. I actually spent less and less time on 3.091 each week. Learning the homework solutions and skimming the textbook turned out to be enough for getting a decent grade, especially since all quizzes and exams were open-book this semester. I also hate organic chemistry with passion. Considering 3.091 publishes grade cutoffs at the beginning of the term and freshman fall grades don’t actually show up on transcripts, I decided to go into the final exam without any preparation13. I ended up recording only 43 hours total, averaging at 3-ish hours per week.

I recorded a total of 466 hours on classes this semester (including the dropped ones) with the median of 33 hours on a single week. I probably spent like an hour or two more every week if you counted the time it took for me to plan out my assignments, manage things on Notion, and do other administrative stuff. It amazes me just how much I’ve learned in a semester, even though it’s just through four classes. (Oh! I also just released my lecture notes for 6.854.)

I certainly wasn’t consistent. Instead of working 6 hours/day, it was more like 2 hours/day and one spike for 6.854 every week. Yet, I still managed to feel as tired as some other people I know spent 40-50 hours per week on classes. I think it just shows that there’s a huge mental component to how draining things get. If there’s anything I’ve learned from this semester, it would be to follow my energy patterns more. (If you know you’re not gonna start the pset until two days before the deadline, why bother agonizing over it for the entire week every week?) Ironically, committing to more things in a day also helped me feel better. It makes sense: When you’re stressed out about work, the best thing isn’t to do nothing and worry about not making progress. Hobbies like gaming or writing (which might seem demanding) are actually a pretty good escape from intense college life.

Notion is aesthetically pleasing, so I thought I should put the screenshot here to fill the space before the next section.

Social Media Detox

This is somewhat related to the previous section, and this is sort of a rant.

I feel social media takes a lot of the time I could’ve used to finish my work or pursue other hobbies. Facebook, Twitter, Instagram, and Youtube are fine when there are genuinely interesting contents. But when you get to the point where you exhaust everything they have to offer for the day and you still feel the need to pick your phone up every few minutes to check if there’s been an update, that’s problematic.

It’s like a toxic relationship: I know objectively I should quit yet for some reason I always find a way to rationalize sticking with it. For example, I tried to create an elaborate plan to ensure that I still don’t miss out on things while not wasting too much time. That never worked. In a way, social media has become a way for people to cope with not knowing what to do. They’ve evolved so much to capture people’s attention. It definitely worked on me.

The short-term solution turns out to be quite simple: Shut everything down as soon as I realize I’m spending too much time on it. Bore myself by doing nothing. In four or five minutes, I’ll gain the clarity that allows me to start working on things long overdue. Of course, this doesn’t always work, but it has worked often enough to convince me that social media are indeed designed to be addictive. CGP Grey articulated this point very well in his video which I encourage you to watch when you can.

Though the issue of social media addiction has been on my mind for a very long time, I’ve only been able to act this past mid-November. Some things happened that led me to think: What if I just go cold turkey? On the spur of the moment, I decided to delete all social media apps from my phone and downloaded an extension to block the websites. That worked quite well, especially since I was busy with final projects and training camps in November14 anyway.

I gradually added back some stuff, but I think I’ve been using social media a lot more healthily than before. I’ve completely cut my ties with Instagram, though, because I realized it doesn’t actually add any entertainment value at all.

The problem isn’t gone yet, though. Turns out there are infinite number of ways to procrastinate, including but not limited to, Discord, Piazza (yes, 6.AcAd is surprisingly fun), Firehose, Courseroad, and YouTube (which for some reason I haven’t removed yet). Guess this will go on for a while.

IAP and spring semester plans

Following the grueling Fall semester at MIT, there’s a one-month semester called the Independent Activities Period (IAP) which is meant to be a short but fun opportunity to do things. For example, you can take web.lab which is simultaneously a crash course on web programming and a competition (with prize pools!) or Battlecode which teaches you to write bots that compete in a real-time strategy game with other people’s bots. There are also some of the more conventional classes, like Calculus (18.02A, extended from Fall semester), Classical Mechanics II, accelerated language classes, and other special topics in math/computer science/etc. IAP is also an opportunity for students to focus on doing research and other activities (as the name suggested). It’s also possible to simply use IAP as an extended winter break before the spring term starts again.

I guess this means there will be another review blog in a month! I’m planning to take Pokerbots Competition and 6.S191 Introduction to Deep Learning. The former just seems fun, yet light enough to offset some of the commitments I have outside of school. The latter, I assume, is a somewhat application-focused introduction to machine learning. I’ve already learned a lot from 6.036 but I haven’t had any experience working with libraries yet, so I figure 6.S191 would be a pretty enlightening experience.

Obama was “invited” to introduce people to 6.S191. Of course, the audio and video were actually synthesized. Not sure how good 6.S191 will actually be, but starting with this is pretty cool I guess.

IAP classes are known to be very intense, but they are graded P/D/F (pass/D/fail), and for freshmen specifically, P/NR (pass/no record). So, the pressure will be quite low regardless. Furthermore, I’m quite lucky Pokerbots lectures are scheduled during the first half of IAP meanwhile 6.S191 is in the second half, so that’ll help spread out the workload. If everything fails, however, dropping a class is also an option. We’ll see how that goes.

MIT actually invites all freshmen, sophomores, and juniors on campus this spring, albeit with some rather tight restrictions (e.g. COVID-19 tests two times a week, strict social distancing, almost all classes still virtual). I know a lot of people are going. Sadly, after weighing all the benefits and headache-inducing logistical issues, I’ve decided not to go on campus. This means I’ll pursue spring courseload entirely online and won’t get to meet MIT friends in person. Thankfully, I think I’ve handled fall semester well, so I can probably tough it out for another semester. I’ve come up with a lot of stuff to do as well, so it will still be somewhat exciting. I’m definitely going on campus in Fall 2021, though, unless the pandemic somehow worsens. I’m somewhat optimistic it can’t get much worse than what we have right now. But we’ll see!

Happy end of the year. Hope 2021 will be a somewhat less exciting year!

A random photo of MIT in the winter (source). There’s generally more snow than this.

  1. Source. See Paul’s blog on MIT slangs (a pretty old one but still relevant). See Slug Wiki. (Do people use this anymore?) 

  2. Notion is a very powerful productivity web app that allows pretty much anything you can think of: calendars, to-do list, journaling, etc. I set up a dashboard that allows me to keep track of my assignments and compile class information so I don’t have to hunt them down on Canvas and hundreds of other websites every time I need something. If you’re interested in learning to use Notion, I suggest Thomas Frank’s video. He explained the features of Notion pretty well. Though you most likely won’t use Notion the same way he does, I’m sure you’ll get a lot of ideas on how to create your own system. 

  3. General Institute Requirement (GIR). Science GIRs include two math classes, two physics classes, one chemistry class, and one biology class. 3.091 is chemistry. 

  4. CI-H: also a synonym for ``you have to write a lot of essays.” 

  5. 256: 2 to the power of 8. Beautiful. 

  6. I heard there are courses with worse organizations, though. Who knows if I’ll end up loving 6.854’s administration? 

  7. Oh yea. It was definitely unhealthy. The total time (minus the proposal) came out to be ~30 hours, so it was about 6 hours a day on average. That might not seem like much, but remember this is 6 recorded hours, when I was focused on actually reading the research and trying to write things. Add in a few more hours of simply dying inside and two whole weeks of not seeing sunlight. Brain fog is real. 

  8. Karger was actually pretty lenient with extensions. Formally however, classes are not allowed to have anything due after the last day of class (i.e. the formal due date for the project). 

  9. Surprising to many other people I know at least. I think this was actually completely expected. 

  10. I only discovered after a few weeks in that I could ignore the readings and only read right before the listening quizzes. Oops. 

  11. Okay, this was a hottake, but at least it’s the impression I got after seeing people complain about 14.01’s readings (non-CI-H, mind you) and 24.900’s final paper. 

  12. Additional info for 21M.011: There are three essays. The first and the last one are analytical essays of 1000 words each. The second one is 2100 words with revision cutting down to 1600 words and is more personal in nature. There are also three “concert reports.” In a normal semester, you would attend a concert and then write a 300-to-500-word reflection on it. This semester, we were allowed to watch recorded concerts from any online libraries. I don’t know what you think about this, but I thought this was the lightest essay workload I could ask for in a CI-H class. 

  13. Confession: I actually submitted a blank exam. I couldn’t bother. I learned later that 40% or so of the class also didn’t care about the final exam. Guess I shouldn’t feel guilty. 

  14. Hmm, now that I think about it, a lot happened in November. The question is: What really happened?