update

short post

lots of good content lately:

• finished before you suffocate on your own fool self, a really well-written short story collection mostly about growing up african-american / mixed-race. now i’m reading the design of everyday things, which is decent but feels a bit repetitive (it contains a few design principles, along with lots and lots of examples)

• watched ocean waves and brokeback mountain. ocean waves is a relatively obscure studio ghibli movie from 1993; the plot is a bit silly (lots of awkward high school drama) but the animation and music are good. brokeback mountain is a pretty sad movie about two cowboys’ relationship; it’s the first thing i’ve watched in a long time that moved slowly without feeling boring. i also finally got around to making a letterboxd (basically goodreads for movies) to organize what i’ve watched, but it’s still a work in progress as i have a lot of movies to populate

---

went to iceland for a week!

• best driving environment i’ve ever had, because the roads are well-maintained and low-traffic and the scenery is gorgeous everywhere you go :) at almost any point in time you’re either passing through interesting architecture or really pretty nature, so it’s hard to get bored while driving. i also drove through my first snowstorm, and it was super cool to see the horizon disappear because the sky and ground were both white

• didn’t get a great view of northern lights unfortunately :(

• my favorite parts of the trip were hiking a glacier and spending time in a reykjavik bookstore

  • if there were no problems to solve in the world i think i’d enjoy being a glacier hike guide; you get to explore some interesting territory and see really nice views on a daily basis

  • the bookstore had some of the best interior design and planning i’ve ever seen; it’s a bar + bookstore + music venue combined into one space, with performances every night

---

i feel like i’ve had the same conversation about ai five times in the past month so i’m just going to record my main thoughts here and start referring people to this post when they ask what my thoughts are

• i’m not really concerned (for the foreseeable future) about risk from ai models becoming sentient; most of the risk i find threatening is malicious humans using non-sentient ai for applications like disinformation (fake news / deepfakes / viral memes), censorship (openai/other corps already have the ability to alter historical facts and most people will just accept those revisions without questioning them), manipulation (think ai partner/dating apps like replika, which people have already reported falling in love with, except much more advanced)

• i see a lot of discussion along the lines of “how do we prevent china from developing agi before us” and, idk, i’m probably biased because i’m chinese, but i can’t help but feel that this is at least partially / maybe mostly motivated by xenophobia? i also think it’s strange that a lot of the people concerned about chinese ai are longtermist ex-risk people since it’s pretty likely given how dysfunctional we are that the chinese government would do a better job of regulating / managing problematic ai than western governments. so yeah i’m not really worried about this either, and i’m confused by how often it gets brought up

• on the other hand i am concerned about wealth accumulation from ai. i think it’s not unlikely that ai continues the tech trend of increasing wealth inequality and/or escalates it in a pretty severe way by creating a few trillionaires and leaving most people worse off. the ai-specific solution here probably involves profitable ai companies compensating data sources so that wikipedia / random writers and artists on the internet / etc are rewarded for contributing to model training. whenever i bring this up my friends tell me this will never work and the right thing to do is general economic restructuring / something like ubi, which makes sense but i don’t feel qualified to write about this