The personal projects portion of my standing plan expects a continued focus on the donations list website (DLW) and migration-related stuff for the rest of 2021. Thinking about the next projects I would like to take up as I get these two to a logical point, I am leaning toward focusing on one or more of the subject wikis (subwiki).
Focused effort on any single subwiki would likely equate to one project, so given my heuristic of juggling two projects at a time, I expect to pick two subject wikis at a time. It's also possible that I'll conceive of a project of the nature of "fun mathematical subwiki editing" that is allowed to hop around across multiple subject wikis all within one project.
My general thinking has evolved quite a bit since I first started work on subwiki, so I need to think a bit more about the motivation and mechanics of returning to subwiki.
When I first started graduate school, for the first few years, I spent a lot of time on subwiki. The vast majority of that time was spent on Groupprops, but I also spent time developing Calculus, Topospaces, Commalg, Market, Mech, and a few others. Groupprops related closely to my research interest in group theory. The others were motivated by curiosity and interest in the subject matter. Pedagogical motivations played a role in my focus on Calculus.
Working on subwiki felt exhilerating and rewarding. Think of the kind of feeling that a person gets from vigorous physical or mental exercise. It's a bit like that, except that here one also builds something that can be looked back upon -- and built on top of.
My gradual drift away from subwiki happened aroun 2013 onward:
-
Altruistic impact: A big factor was an increasing interest in "high-impact" areas, partly due to the influence of ideas like effective altruism, as well as interest in migration / open borders.
-
Exit from academic mindset: Another factor was my shift away from academia, which meant being less in a mindset of research, pedagogy, and building.
-
Day job offering less continuous time away: A very busy day job played a role. With a busy day job, I had relatively short periods of free time, and subwiki requires more continuous time to allow for immersion. Back when I had been in academia (before finishing my thesis became critical), there were periods of time such as summer vacations when I could just focus on subwiki.
-
Competing activities: I picked up competing activities more aligned with the above, further reducing time for subwiki: Wikipedia editing (more pageviews, easier to fit a diverse range of interests) and, after being booted off Wikipedia, into Timelines Wiki. I also started new projects like the Donations List Website (DLW) and helped on projects such as AI Watch and Org Watch.
None of these areas gave me that feeling of exhileration and reward that subwiki had, though they were satisfying in other ways. I also didn't feel for any of them that it relied on my unique skills and value add to the world to the same extent that subwiki had. Finally, in terms of pageviews, subwiki is still second only to Wikipedia in terms of pageviews, and the subwiki pageviews are less confounded by counterfactuals.
As my work life settles down and I reduce time on work, some of the reasons for drifting away have reversed or at least have become less strong. Let's run through the reasons to understand:
-
Altruistic impact: Back when I was initially interested in areas like migration policy and effective altruism, the domains were still relatively nascent and the intellectual communities relatively small. I had a sense that I could meaningfully contribute with limited time available after work. Now, various communities have matured and/or collapsed, and I feel like my ongoing participation isn't important (though some of my tools are still valuable).
I also am tentatively leaning toward the view that the ability to add unique value and perspective is important, and I feel that I can do so more through subwiki, because I have a perspective there and the work I do helps flesh out that perspective. Considering altruistic impact, though, does affect the selection of subwiki. My original focus on group theory was most aligned with my own interest and passion, but group theory has limited altruistic potential. I will discuss the selection of subject wiki in a later section.
-
Exit from academic mindset: While there isn't much direct change to this, I do feel like I have more of a "pan-academic" mindset now: incorporating the academic virtues of truth-seeking, but not academia's institutional qualities. I feel like I can work with this pan-academic mindset across many kinds of life environments including my present and future ones.
-
Day job offering less continuous time away: I have reduceed work hours at my day job. While I'm currently using the reduced hours in a distributed way (rather than taking long, continuous stretches of time away from work), I expect to be able to take continuous stretches if that turns out to be the bottleneck for subwiki. At minimum, even at present I usually have around 1 day per week where I am free of nontrivial work and nontrivial personal chores, and by arranging the 1 days back-to-back I can get 2 contiguous such days.
-
Competing activities: Some of my competing activities are outsourced -- including a lot of work on Timelines Wiki. Some others I might be winding down or transitioning. Yet others, like Donations List Website, I hope to streamline enough that they'll effectively become low-effort chores. Moreover, with my alternating projects strategy, I can focus on subwiki for at least some time.
There are three pieces I think of around value for each subwiki:
- Personal excitement to me (intellectually and emotionally)
- Appeal to other people in the short term
- Long-term indirect value
Working on Groupprops was personally exciting to me. There were times when the excitement came from the intrinsic beauty of the material, and other times when it felt more like doing exercise and felt satisfying in that sense. Even when it was more rote than beautiful, I still think my familiarity with the underlying basics of the material made it more enjoyable for me.
In general, I expect the personal excitement to me to be higher the closer to mathematics a subject is. There are probably going to be exceptions, though.
How much will people use and enjoy the subwiki? I think of this in terms of four pieces:
- Demand for the stuff to begin with
- Extent of competition
- My unique value-add
- Amount of work I do on the subwiki
Groupprops is by far the subwiki with the most pageviews. It a case of low-to-moderate demand and low competition, which sort of cancel out. The two main things where it stands out are my unique value-add and the amount of work I do on the subwiki.
I think that on many mathematical topics, and maybe also on learning, my unique value-add will be a little lower than for group theory, but still pretty high! The amount of work is the main variable I control. However, many topics, like calculus and machine learning, have way more demand and way more competition, and that information would also need to be factored in.
How does subwiki create altruistic value beyond the short-term value it provides to people? I think mainly of two channels of indirect impact:
-
Value of the domain to progress: Some domains have significant value for the world, and subwiki adds to that value by helping people in those domains.
In contrast, if subwiki is only used by school students to get better in what is mostly a zero-sum signaling contest, it creates very little value for the world.
-
Value to shaping people's imagination: The subwiki style of thinking about topics can unleash people's imaginations, and help them explore and discover related ideas that ultimately makes them better people. For instance, think of things like HPMOR or the LessWrong sequences.
Moderate/high. Calculus definitely isn't as thrilling for me as group theory. But it has the same comfort factor for me in terms of understanding of the basics, which allows me to explore. I also have experience teaching the material, and the calculus subwiki is already started so I can pick it up more easily.
-
Demand for the stuff to begin with: There's a lot of demand. Most of it is driven by school and college curricula.
-
Extent of competition: There is a lot of competing calculus content, much of it more tailored to specific sources of demand (for instance, specific school curricula, problem-solving focus).
-
My unique value add: My experience with calculus pedagogy, various videos I did, as well as a deep understanding of the relationship with other areas of mathematics and other sciences helps me offer a boutique perspective.
-
Amount of work: This is the variable I need to decide. I think about six months with this as one of two non-work projects will be needed to make a meaningful stride.
Let's look at the two channels of value:
-
Value of the domain to progress: Probably not much! Calculus is extremely important, but also well understood. The people looking at this wiki are highly unlikely to use the information they acquire to make the world a better place.
-
Value to shaping people's imagination: I think there's a chance that there is some value here. Many of the people coming across this resource might be students in a zero-sum school or college environment, but the fact that they're even Googling and looking for resources is a (weak) indicator that they are open to new ideas.
My academic involvement with machine learning has been low -- I've been mostly self-taught and picked it up from friends and the Internet. I do have relevant work experience though. Working on the machine learning subwiki will be an opportunity for me to give my understanding a more rigorous foundation.
Despite my lack of academic grounding, I'm actually pretty confident of my baseline understanding, as with calculus and group theory.
I find this exciting but maybe a little less than calculus, mainly because of the slight handwavy nature of the pages.
-
Demand for the stuff to begin with: Demand is intermediate between group theory and calculus. It'll be a mix of college students and existing or wannabe data scientists and data / machine learning engineers.
-
Extent of competition: I think the level of competition is moderate, again in betweeen calculus (where there's a lot) and group theory (where there's very little).
-
My unique value add: Not sure it's super-unique! But I do have some experience with both machine learning and with subwiki, and the nexus can give me some insights. Also, I can develop this along with calculus subwiki (e.g., gradient descent on calculus and gradient descent on machine learning) which allows for a powerful bilingual development of ideas.
-
Amount of work: This is the variable I need to decide. I think about a year with this as one of two non-work projects will be needed to make a meaningful stride. I'm a little farther behind here than with calculus, where I already have a basic infrastructure of content.
Let's look at the two channels of value:
-
Value of the domain to progress: Machine learning is one of the cutting-edge domains affecting progress at current margins. I don't expect this subwiki to immediately get to the point of affecting that cutting edge, but it is more likely to do it there than with, say, calculus.
AI safety concerns raise a question about the sign of machine learning progress. As a person with concerns that are somewhat more safety-aligned than most people (including my focus on interpretability and deep understanding) I expect that the flavor of progress that machine learning subwiki will make more likely will be a bit better for safety than the coounterfactual.
-
Value to shaping people's imagination: I think there is some potential here, but honestly I would need to first shape my own imagination.
Learning holds a "meta" excitement to me -- as I work on the other subwikis, I think about teaching and learning, which feeds back into stuff that I can write on learning subwiki.
Unlike calculus and machine learning, I may be doing a lot of indirct primary, secondary, and tertiary research when writing stuff here -- or just shooting from the hip based on my already-acquired experiences.
-
Demand for the stuff to begin with: Demand is moderate but not huge. A lot of demand would have to be generated by getting people curious about the mechanisms of learning.
-
Extent of competition: There isn't a lot of similarly formatted competition. There's a lot of material on learning but usually in different formats.
-
My unique value add: A lot of my unique value add comes from my pedagogical experience as well as my work on subwiki.
-
Amount of work: I'm guessing six months to a year will be needed.
-
Value of the domain to progress: Mainstream education is unlikely to move at all, but autodidacts and inspired educators can benefit.
-
Value to shaping people's imagination: There's a lot of potential here. But I would need to get plugged in more to get a sense of the potential.
Understanding Econ 101 ideas -- their basic form as well as the nuances surrounding them -- has always excited me intellectually. It's not a mathematical Everest, but it stands in the valley between theory and the real world that gets me excited in a different way.
-
Demand for the stuff to begin with: There's a lot of demand for econ stuff -- coming both from school and college kids and from adults whose interest is kindled by the econ blogosphere.
-
Extent of competition: There's a moderate amount of competition -- but maybe less than for calculus!
-
My unique value add: My unique value add comes from a combination of my grasp for economics, my deep understanding of mathematics, my wide cross-topic reading, and my knowledge of subwiki.
-
Amount of work: I think a six-month period will be good enough to get started.
Let's look at the two channels of value:
-
Value of the domain to progress: A calmer understanding of the mechanisms of Econ 101 can lead to a calmer, more peaceful world. But it's not quite tied to cutting-edge progress.
-
Value to shaping people's imagination: Understanding how to think of the econ/sociology/psychology junction can shape people's imagination on how to understand the world around them in other respects as well.
The topic seems moderately fascinating as a pure mathematical topic, and has a similar appeal as group theory. Unlike group theory, it has a "meta" flavor to it insofar as it describes how we reason. So that makes it more exciting.
I think a lot of the things I did organizationally with groupprops could port over to something on logic / model theory.
-
Demand for the stuff to begin with: The math/theory crowd would likely be a little smaller than the crowd for group theory, mostly because it's less standaard in undergraduatee curricula. But the CS/hobbyist crowd would probably be bigger.
-
Extent of competition: Not sure! But competition in a wiki-like format is rare. Maybe something like n-category cafe. Ultimately Wikipedia.
-
My unique value add: This mostly comes from my past experience with Groupprops leading me to have a picture of how to structure the material.
-
Amount of work: Since I'm starting from scratch, probably about a year!
-
Value of the domain to progress: The main channel is likely an influence on the design of AI and thinking about AI safety. It would be a while before it gets there, though.
-
Value to shaping people's imagination: Similar to groupprops, it has a kind of artistic beauty and helps us be more self-reflective. More so since the topic itself is meta.
Cellbio subwiki, which already exists but is very much in infancy, would be a radical experiment within subwiki. It's me trying to use subwiki page creation as a way of learning a topic I have no demostrated aptitude or intuition for. The main excitement to me is in its utilitarian value of helping me get a better intuition for biology. That may ultimately help me personally, but it's more interesting in terms of helping me understand research on life extension (a lot of which is cellular).
-
Demand for the stuff to begin with: Moderately high as it's in school and college curricula. But unlike calculus I don't think people "struggle" with understanding concepts so they won't rush to look for online resources that much.
-
Extent of competition: Wikipedia is probably pretty good and would be hard to beat in coverage (only in presentation).
-
My unique value add: This hinges on the subwiki presentation format.
-
Amount of work: Since I'm starting from close to scratch and don't know much about this, probably about two years.
-
Value of the domain to progress: High, but I don't think a subwik helps cutting-edge research much.
-
Value to shaping people's imagination: I see this as helpful to amateur consumers of biology research to get their bearings on what's happening at the cellular level. But it's a ways out from that.