this post is heavily based on/talks about this post on sparkly people/community building & also this post on anti-sparkly language/slightly critiques the former (I'll reference the former as the "sparkle post" and the other one as "anti-sparkly post", even though I know the critique is more nuanced than just anti-sparkly)

I love discourse that models the way my brain approaches ideas—sort of like a socratic dialogue between two sides (if you do debate I literally imagine a flow)—and the two articles above model a dilemma I’ve been thinking about for a long time. Starting from the POV of wanting to be incredibly intentional with everything I do and what I am being the amalgamation of the ten people I interact with the most:

Should I optimize the way I interact/meet people? If I should, how do I even do that?

Splitting this up into two questions actually models the way my mind has changed about this from last year to this year. Freshman year I actually aligned heavily with the anti-sparkly post, mainly trying extremely heavily to lean into randomness—everyone is cool, I should expand my circles. Given at this point I wanted to “expand my friend group outside tech-bros”, I think this was a really good vibe for me to lean into.

This functionally starts from the idea that “all people are interesting”. Up to this point I agree with the anti-sparkly post. Functionally I believe that if I don’t find someone interesting, it’s actually my fault. All people have unique life stories. This should probably be an axiom in whatever conversation you’re having, and it should be up to you to uncover what the story is. This is the approach I took in my first year—meet tons of new people you never would have met before.

Although this worked, it led to a feeling of ennui and “unproductiveness”. Every conversation I had was about new topics—music, history, robotics—things I had never thought about. On the flip side this led to almost no conversations about my passions. I couldn’t talk about the things I loved, or the topics that took over my life—any project I was working on got relegated to the sidelines. Why? Although I did get a giant broad array of conversations, I was always handicapped by my lack of knowledge in the area. I am pretty good at learning fast and asking the right questions, so this usually wasn’t a huge problem—this wasn’t always reciprocated though (I doubt lack of ability or “intelligence” is a reason here, its probably a variety of factors, people can’t always put their 1000% into conversations). The addition I realized that should exist is “All people are interesting BUT I should interact with people that help me grow as a person”. I can interact with and find people interesting, but still never improve as an AI researcher & stay at the same skill level.

Interacting with other people/broadening your horizons is important, but it shouldn’t fundamentally take away from your interests as a person. A good point of clarification is that this “random walk” of finding completely random people interesting, does not work if you have goals. You could constantly talk about yourself and your own goals, but you’re bound to scare people away/have people find your own topics uninteresting.

I think the anti-sparkly post fails to understand something fundamental about meeting new people. The people you interact with change you. You should be intentional about how you change. You’re probably optimizing for “something” subconsciously, and I think its better to probably lay the terms out concretely so you can de-bias yourself and check your gut instinct. The part the sparkly post gets a bit off is having too many heuristics (theres no way I can run down this list fully when meeting people), as well as not having a great basis for why we need “sparkly” people—it sort-of comes off as elitist if you’re not in the niche community it was meant to be a part of.

So… whats the solution here? I think vague heuristics in general are probably bad here because of reinforcing biases and getting stuck in loops. Also I’m not a fan of having a bunch of heuristics because I probably won’t end up using them/usually there has to be a simpler explanation. After thinking a while here are the couple traits to focus on when meeting new people/deciding who to spend time with. If I had to specify my objectives, they would be (of course with added stochasticity so I don’t reach a minima):

1. Are they able to ask the right questions?: I feel like this is a proxy for a lot of different things—do they accept things blindly? can they update their world models fast? are they able to roleplay different ideas/be able to be neutral? By being able to ask hard, interesting questions, it expresses a lot about the person. This by far is the hardest thing to find empirically. Many people just accept explanations & don’t add on, or don’t have the background knowledge to actually contribute to a lot of conversations. I think 1) this shouldn’t reflect badly on the person, some people might just not know a lot about the area, but 2) it doesn’t make it bad to optimize for—I want to interact with people who are able to adapt to different scenarios, finding them interesting, and asking the right questions.

   This might seem unfair but I’m also including offline conversations here—texting/book recs/etc. Can they actually ask questions/think about recommendations and offer good criticism + praise in contexts that are wildly different than their normal areas of research?

2. Do they make?: Here I like to think of it in the broadest aspect—do they like to do things? This could be literally anything from crocheting to building racecars. As long as they have a passion and create anything, I feel like this one is checked.

3. Are they kind?: This by far I think is the easiest and most common. Interacting with unkind people just isn’t fun. I think the post below does a really good at disambiguating kindness and niceness. Kindness seems like a less superficial quality in a sense? I’m definitely fine if people are not nice, sure having a bad day is normal—but being kind is an important trait (arguably also this is a proxy for rationality, most rational agents should cooperate because of the repeating prisoners dilemma).
   
   A great way to approximate this is “how iroh-like” is this person. Even if that person isn’t having a great day, they ideally would go out of their way to help their friend, or hit the cooperate button. The wiser and more “iroh-like” the better. I think this is important because someone can seem sparkly just because they’re nice, but its harder to determine if someone is kind. Being able to determine this entails being able to make third party neutral observations on the person and how they act in different situations.
   

#42: Are you nice or kind?

Last week Avi and I went outside to unbury our car after the storm, a task we assumed we could do ourselves but ultimately involved our whole

https://haleynahman.substack.com/p/42-are-you-nice-or-kind?s=r&c...

4. (OPTIONAL) I think this should be optional because this is very hard to do and might not actually be that important… but I think on they should be on the same scope as me (heres a backlink to the blog) in their view of the world.

   “Most games in life are repeating (ie. your life won't end if you fail) yet people are still competitive. Why? I think the answer is scope. Most people forget the global scope that is key to gaining perspective and achieving what you want.”
   
   ”…some people get locked into an extremely small scope of what they can change in the world. If you only think locally, then of course the only change you can have is local. A global scope encompassing over as much as possible, is necessary to have higher levels of influence.”
   

   I think that putting thought into the scopes that you evaluate life with, as well as evaluating abstractions is not only good, but pretty much essential if you want to be happy. Its similar to wanting to be friends with people in the same stage of life as you (ie. I wouldn’t find much in common with a kindergartener) but in a much smaller way. Its a question of prioritization, if your scope of change is the whole world—trying to change the political system—and another person’s scope is limited to college & the next grade on a test, it might be hard to find common topics to care about. Though its easy to maintain conversation, there would always be a slight disconnect.

There we go! Three (to four) things I think are really important for me to optimize for in meeting new people. Being able to ask the right questions & making are the couple I’m focusing on right now—can I have a discussion about my research & still have the conversation be interesting/have new ideas thrown into the arena. Believe it or not, I actually think a lot of people fit the 3rd category and the 2nd (so many people have passions that are centered around creation). Different ways I’m trying to optimize for the first include (but are not limited to)—joining new areas which force people to learn fast, hack nights, and being much more open about my research! Hopefully by ranting about different things I find interesting, I can self select for those people! Zooming all the way back, this is just a heuristic for making friends that make you happy—I try to make it a goal to not confuse abstractions with the goal itself. If at any point abstractions stop you from meeting cool people, the abstraction itself is wrong.

Hopefully this was (pretty?) interesting, this is a push in my life to be much more intentional about what I do. This push isn’t recent, I’ve been trying to be more intentional over the course of the past 3-4 years, but of course I can always get better. Being intentional with friends is honestly amazing, I think its a pretty simple thing to look into that has large effects on your day to day. Next up… life updates!!

P.S: Not intentionally being friends with someone != to someone being your enemy. I still believe in the quote, “You have no enemies. No one in the world is your enemy. There is no one you need to hurt.”

a couple much needed updates:

• back @ UCLA! I’m spending my time doing 2 days of independent research + 3 days of school + 1 day hiking! Its been going well so far, I really enjoy the hiking bit because I feel like @ UCLA I can get “lost in the sauce” on campus a lot. Exploring is fun.

  

  

  

• currently doing independent research! here are a couple ideas on the cooking plate—
  • LoRA layers + reasoning, can they significantly improve reasoning?
  • similar to the ROME (Rank-One Model Editing) paper https://arxiv.org/abs/2202.05262—can we identify where world knowledge vs reasoning are stored?
  • 🥫mixture models + continual learning

• dumpster diving updates!!!
  • my room has a 3d printer & a typewriter now—both found and refurbished! The 3d printer wouldn’t print and the typewriter wouldn’t type, and after finicking with both, I have a cool little workstation!

  

  

• was in SF over the summer & worked @ Notion/did autoregressive audio research @ stdint this summer! Was really fun, will probably dedicate a whole post to what I learnt
  • TLDR; I love corporate life, I think focusing on one thing is actually really helpful vs school where I have to balance a lot of different aspects
  • I also fixed a ton of stuff @ notion and worked on a bunch of interesting technical problems (latency, web search, markdown parsing, etc.)
  • living alone in san fransisco was AMAZING, I love the freedom—only thing I would change is maybe strike a balance between living alone + dorm life

• had a side project on diffusion transformers from scratch scratche-dit pt.1 (DiT implementation in ~130 lines) to upskill on doing harder math in research papers

• re-vamped my dotfiles! https://github.com/neelr/dotfiles

• climbed half dome!!

goals:

• old: make new friends/find as many different people as possible ⇒ new: focus as much on research as possible + talk about it more openly with people

• old: look for different labs + organizations to do research with ⇒ new: experiment fast and have structured time to work on your projects—also write more and document the work that is done