poast office ii

food for thought

• why is trust important? how do we create it? how do we break it? there are lots of fascinating intersections

  • game theory (repeat games; reneging)

  • history (ancient and modern)

  • psychology / biology (we feel good when we act trustworthy)

• what can we do once trust exists?

• think about something as discrete as an auditing systems. auditing has enabled so much trust in difficult-to-verify domains! think earnings reports for all publicly traded companies. what an incredible invention

• consider monetary policies: why does the united states dollar have any exchange value at all? how truly robust or surprisingly delicate is the value of the usd. trust in contracts and systems underpins a lot of this. how anti-fragile are those systems / institutions? is there a reasonable future where the value of usd → zero? or is it ‘too big to fail’?

• relatedly, i've thought about how you build, destroy, and maintain trust in the ai 2040 proposal

  • if you make a deal with your competitor, how do you design the agreement so all players are maximally incentive-aligned

    • i.e. “i don’t want to break this deal, because i would be in a worse of position if i did that”

  • in ai 2040, there may be baked-in assumptions, like people agree a world-destroying arms race is bad for everyone (e.g. nuclear)

  • does this work for something as single-breach sensitive as superintelligence? one leak and you’re donezo?

    • on first pass, it does seem similar to nuclear or bioweapons

      • one nuclear code leaked to the wrong person could lead to global catastrophe, even if politicians in each country are level-headed and understand anomaly mistakes

      • one leaked virus could lead to a global pandemic

    • if that’s the case, we already have international cooperation norms and policies that we can use as models (not that kind) for future deals

  • what are the ways superintelligence might be different?

  • said differently: why wouldn’t we be able to make an international agreement between competitor countries on ai deployment and safety infrastructure?

    • is it truly insurmountable?

    • or is a deal inevitable?

    • surely we will rally around the seriousness of cooperation if we see the realized risks? why do you disbelieve this?

  • what are the ways the nuclear / biorisk models don’t transfer to superintelligence?

• communication is really expensive. verbose communication even moreso. some of the most effective leaders i know say in 10 seconds what others say in 100

• did the outcome change with more words? if not, are you communicating the way you desire to?

• often there is a tension between precision and brevity. the details often matter! or at least it feels like they do. but do they?

  • you may want to say: i 60:40 prefer chocolate:strawberry ice cream

  • and want to provide more details: i prefer chocolate, but i just had chocolate the other day at bi-rite, and anyway i heard this place specializes in sorbets, and i picked the coffee this morning and i want to be fair and... so i defer to you

• there may also be a tension between subtext and brevity. i want to demonstrate modesty, deference, etc.

  • if you want to communicate: we will not be able to deliver the project on time

  • you may soften it to your customer who is in a position of power: sorry to bother you, but we have an unfortunate note to share that we… won’t be meeting our desired timeline and volume 

• there may be tension between communicating a precise feeling and brevity. thoughtfulness + depth vs brevity. domain expertise, and so on…

• but! you can write less and convey the desired message

• the thing i really want to talk about is the meaning of language. can we really ever understand each other? if i say, i love strawberries, that’s different from i love my mom, and is different from someone who really loves strawberries and eats them everyday. i really just like strawberries, actually

• what about red? strawberries are red, but when we look at the same object how can we know my red looks like your red — really?

• wittgenstein would have something to say about this, as would max cooper:

• finally, i’m interested in how precise language affects strategy. if think from a higher plane, you will be running towards the right goal

• for another post: why do we feel so good when we communicate well? why do we feel so bad when we communicate poorly? what motivates us to be understood?

  • i’m probably reaching for some biological explanation here

• for another post: why do some people have creative execution dialed up and others have creative consumption dialed up; compulsive artists vs encyclopedic consumers
  

• i’m doing a bad job of being young. am i going to regret this?

• 1) what i really mean to say is i’m doing a bad job of enjoying my youth, while i still have it. i struggle to prioritize time for concerts, or grabbing drinks / dinner with friends, enjoying park days and hikes. reading books, learning something fun. learning while your young seems especially sweet when you can interact with the world in reference to other things you know. e.g.:

  • fred again used to be in brian eno’s choir, so he’s heavily influenced by the ambient and generative music eno pioneered. i think about that every time i listen to a fred again song, which is often

• 2) is this a thought others have had and come out on the other side of? what advice do you have if so?
  

• dealmaking in DC: someone recently described what it’s like to make deals in DC. the majority of convos are at happy hours / bars, shooting the shit, and gossiping, with just a fraction of the in-person time actually spent talking about deal logistics / world views / philosophical debates

• i’m curious if, related to the effective comms section^, the best ways to make deals are

  • mostly have fun

  • info exchanging (novelty-focused; gossip)

  • minimal persuasion / business

• the connection i made to my own life is that this could be a way to think about writing. we have a million random thoughts, many fun, many light-hearted, some curious, many humorous, some serious. this week, this post is mostly fun / curious with some items for serious consideration (ai 2040; oai <> huggingface blog). how can i turn my attention to important / serious things and maintain delight / curiosity? like this, i hope

• another analogy to the DC dealmaking convo: it seems like the skillset of a highly functioning LP of a small / angel fund <> donation managers of political campaigns. it’s a high-leverage service to be the funnel between money + capital + status <> boots on the ground (startups or policymakers)

tidbits // poasts

• maybe b. f. skinner and the behaviorists are onto something when they say “i know that was good for you, how was that for me?”

• there’s something really rewarding to receive positive feedback and makes you more likely to rate that same interaction as positive. sorry i mean this in “i had fun, how about you?” and me feeling like “actually, you’re so right. that was so fun!” when before i was neutral or undecided

• strange how such a small action can be powerful. like texting all your guests after a party to say how lovely it was to see them

• friend who does not live in sf was shocked to learn that there are group houses with people across basically layer of the ai stack. one house: openai, anthropic, nvidia, neo lab, ai regulator + neighbors at deepmind + caisi + overlapping ai policy circles in dc

• that’s just sf, man
  

• overheard in SF: i’m pilled on the codex app now, i just converted from codex cli

  • me: (‘⊙_⊙) (っ º - º ς) just now…?
    

• i love the idea of sending your future self letters / emails where you talk about what you’re paying attention to right now. what beliefs you have, what problems you’re working on, and what predictions you have for the future. i think it’s so delightful to receive a letter from yourself in the past. i have a running list of questions (one for each day of the year) and i’ll share a few of my most recent favorites. i imagine i’ll love re-reading my answers to some of these in the future

  • what was the last thing that made you laugh really hard?

  • what historical event have you already experienced that people in the future will ask you about (e.g. what was it like to live through covid??)

  • what is your current favorite word (or piece of slang) and why?

  • who should you write a thank you note to today?

on my reading list

articles

• ai 2040 - ai futures project; on many people’s mind with the launch last week

• 50 things i know – i really love this article or what it points to. that there are known truths about the world, from your pov, that you’ve learned over this brief and wondrous experience of life. i’ve challenged myself to come up with 20 things i know by the end of the year, and i plan on sharing it

• looking for alice - i also really love this article. i share it with friends when they’re unsure of what to look for in love

• openai <> huggingface security incident blog