The Exponential Returns of Working

The one thing Stanford taught me was to work. The one thing I learnt at Stanford was that working is hard. Considering the fact that this is me writing this in the flesh, I’m alive. Which means, working (hard) didn’t kill me. However, I truly believed I was going to die with the piling work assigned to us. First of all, COVID brought too many new challenges. No more in-person classes, no exams, no meeting people for projects, no standing in line for getting questions answered. Things changed. Everything was on Zoom – classes, project meetings and TA discussions. What couldn’t be on Zoom was put out in a new way. Exams were made longer, and more application-based. Assignments number went up from 2-3 per quarter to one a week. This one change made the difference between studying to learning.

Surprising things happen when you do something over and over and pay attention to what you’re doing. You notice patterns. Things which repeat, and things that don’t. What do you do if you have an exam soon? You (hopefully) study for it. If you don’t have one? You (mostly) don’t. What if you had no exams? Only assignments? Or an assignment instead of the exam a.k.a. the take-home exam? All these were not only honor-based but also effort-based. You could just copy answers from your friends. If you had friends… or work on them till you get an answer. And when you have assignments all the time, you are studying and doing the assignments all the time. This repetitive thought and effort based system led to an increase in 1. Stress (duh) 2. Understanding of the subject and 3. Comfort with the subject. Let’s put these three in context.

Have you ever had the feeling where you get your answer sheet and realise you’ve made countless stupid and avoidable mistakes. And some questions you didn’t know the answers to were in post-exam reality, solvable? I do this all the time. Why? Cue to the three points I mentioned before. I get stressed before the exam, forget many answers and make simple mistakes. However, by the end of the final exams, I get very familiar with the subject. And thanks to having 6 sets of exams in a year when I was young, and the applicability of the topics I learnt, even though I didn’t do much homework, I remember “Mitochondria is the powerhouse of the cell” (HAHAHA).

Moving on to college, the assignments are sparse, classes held just twice a week and there’s not much feedback with learning and application. But with COVID, it was back-to-school time. That’s when I noticed something with the assignments suspenseful music plays. Consider an assignment for 100 points. There would be 5 sections in the assignment with a point distribution like (5, 10, 30, 40, 15) points for cleaning the data, processing the data, creating a simple model, creating a complex model and, deriving and plotting the results. You would guess the time and effort distribution would match the points for each section. However, the distribution would look like this (30, 15, 30, 20, 5). Pretty skewed right? And when it took 10 hours for getting the first 15 points, I assumed I would never finish it on time. As I kept going though, the other steps would get over real fast. The reason being the next sections were dependent on the first ones, so the actual effort of each section was 5, 5 (10-5), 15 (30-10-5), 10 (40-30/2-10-5), 0 (15-40) [footnote for calc]. Now it looks kinda similar to the time distribution right??????

30 15 30 20 05 <- Time distribution

05 05 15 10 00 <- Effort
05 05 05 05 05 <- report
20 05 10 5 00 <- familiarity
30 15 30 25 05

Not really… You would say the first two sections still need way more time what the differential says. I would say you’re absolutely right. However, it’s not just the sections depending on each other, but also your skill. A new assignment usually means you have no familiarity with the dataset or the problem statement. This leads to a lot of the initial time being lost just for being familiar with it. Once you are, the delta you need to learn for each new algorithm is almost zero. So, let’s take time for familiarity (20, 5, 10, 10, 0) and time you need to write the report (5, 5, 5, 5, 5) into consideration when we add up the assumed effort by the professor from the previous section (5, 5, 15, 10, 0), we get (30, 15, 30, 25, 5). Doesn’t really add up to a 100. My opinion is that the extra 5 points here are due to wasted effort (or some more report writing). But the entire point is

  1. Professors gave the weightage exactly based on the importance of the problem!
  2. It’s just that cumulative effort is important
  3. Time is directly proportional to effort [footnote].
  4. Familiarity is something I should’ve considered. I’ve always considered manual labor a part of effort but even getting familiar with something (or someone!) takes the greatest effort.

We’ve seen the factors at play. What to do about it?

  1. Watch someone else do it first till you’re comfortable.
  2. Learn the fundamentals first before trying out complicated tasks.
  3. Take your time, don’t stress.
  4. Follow along while watching. Don’t do it by yourself lest you learn bad habits.
  5. Try combinations and tests you’ve created using the fundamentals you learnt.

P.S. I totally made up the numbers independently and they actually just added up (almost) correctly! I’m actually pretty shocked and also excited about this fact.

P.P.S. The equations:

  1. Point/Importance Distribution -> (5, 10, 30, 40, 15)
  2. My Effort distribution -> (30, 15, 30, 20, 5)
  3. Section dependencies were: 2 <- 1; 3 <- 1, 2; 4 <- 1/2(approx) * 3, 2, 1; 5 <- 1, 2, 3, 4.
  4. Normalized Effort Distribution -> (5, 5, 15, 10, 0) {1-2 based on 3}
  5. Report Effort Distribution -> (5, 5, 5, 5, 5)
  6. Familiarity Effort Distribution -> (20, 5, 10, 5, 0)
  7. Total Effort Distribution -> (30, 15, 30, 25, 5) {4 + 5 + 6}
  8. (2) ~ (7)