I will not trade in this market.
Definition of "finish": I plan to watch all of the lectures, consume enough supplemental material to genuinely understand the material, and do all parts of the projects that are not gated behind actually being enrolled in the course. If something is gated and I can find some substitute that is available to me (e.g. renting H100s at retail instead of using the class cluster reserved for actual students), I will use that instead.
Background: CS undergrad degree, ~5yoe in industry. Current role is moderately demanding and related to ML. I've had rigorous exposure to ML in school and have done some reading about transformers, LLMs, etc in the past.
To further motivate myself, I commit to donate $100 USD to a cause/organization/etc chosen by largest NO holder in the event that this market resolves to NO. If there is a tie, I'll do something reasonable like a fair coin flip between the contenders.
I will post weekly progress updates.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ49 | |
| 2 | Ṁ39 | |
| 3 | Ṁ17 | |
| 4 | Ṁ13 | |
| 5 | Ṁ0 |
People are also trading
The largest NO holder at resolution (sorted by position, descending) was @evan . I will reach out to them via DM about the personal motivation component of this market.
Sorry about falling behind on updates. I’ve made it a bit further into the course, but (as traders have already surmised) there’s ~0 chance of me finishing within the next 4 hours, so I’ll go ahead and resolve.
Reflections:
A set of contingent personal (65%) and work (35%) obligations prevented me from spending as much time on this as I would have liked. The personal component is largely downstream of decisions I made some years back, and I am seriously considering making some significant life changes in the next ~2 months to let me focus better on projects like this one.
The assignments ended up being more substantial than I anticipated, and some of them had finicky performance/timing requirements that took a disproportionate amount of time (relative to just understanding the conceptual stuff) to get right. In retrospect maybe I should have timeboxed these harder and allowed myself to proceed with merely correct implementations, but I thought that might be contrary to the spirit of my commitment. Will think harder about the wording of these things in the future.
Week 3 update: still working on assignment 1 (boo).
Some additional colour: the assignment has turned out to be a lot more work than I anticipated (building a BPE tokenizer and transformer from ~scratch, with nontrivial performance requirements). I also forgot how much slower coding without AI is. Will try to knuckle down and finish it this week.