do you have an ambitious idea for how to make AGI go well? do you need money? do you hate bureaucracy and friction? apply now for microgrants!
please read the entire doc before applying. only send applications to the designated location, or they will be automatically rejected.
would people find an alignment microgrant program useful? handing out grants on the order of $10k with very minimal bureaucracy—a 15 minute application process and a 30 minute retrospective call 3 months down the line. is this a quantity of money that enough to matter at all? is overhead around grant applications actually a big issue? is there really good work that nobody is funding already?
i want to spend money on making there be more good alignment researchers and care about them actually caring about alignment. what would you recommend i do?
maybe i’m overconfident but i generally find that it’s somewhat easy to tell if someone genuinely cares. you can’t really fake genuinely caring (or at least as genuinely as one can about anything—perhaps everything bottoms out in some other drives, but that’s good enough for me). there are some false negatives—people who genuinely care who have taken on the affectations of faking it, because they mistakenly think this helps their chances at success. there are also some people who care so much that they go crazy and start being counterproductive. but i don’t think i’ve ever heavily overestimated how much someone cares about AI safety.
this made me remember something: i feel like the vibes of lighthaven would be improved a lot by having real grass. i’m aware this would be more logistically difficult but i think it would be worth it.
it feels like jirgendwo should be a german word. an antonym to nirgendwo
a fundamental divide
in the domain of things, the best way to accomplish anything is usually to aim directly at it. it doesn’t matter how virtuous you are, or how well intentioned your inquiry is, if you are not aiming directly at the thing. reality doesn’t look on your virtue and reward you for the virtue per se. if you’re bad at something, you practice until you’re good at it.
in the domain of people, the best way usually involves some amount of being virtuous and doing things for the right reason. if you lack friends, it’s cringe and comes off as desperate/needy and even manipulative if you optimize too hard for meeting people and how to induce emotional connection and trust. rather, you should work to become a virtuous person who genuinely cares about people and then things will work naturally.
i think this leads to a lot of miscommunication between the groups. it seems desirable to develop the ability to do both depend on the setting.
does generous behavior in abundant situations tell you very much about someone’s character wrt kindness, generosity, etc?are people who are ungenerous also stingy when the personal cost is very small?
let’s run the actual experiment and see? shouldn’t be too hard.
obviously the cot semantic drift will initially be similar to language. but it will only get less legible from here. and this is already pretty bad? i wouldn’t be surprised if two different people tasked with independently deciphering this would come to very different conclusions.
obviously the glorious international pause treaty is a good idea, and i think people have always overestimated the cost of saying so publicly.
the way you do coordination is by saying you would like X if only everyone else also wanted X. you ease yourself into it, making tiny steps and ceding imperceptibly small pieces of ground until coordination has already been accomplished
oh, there’s no reason to expect evolution to have given us the necessary tools at all. certainly i’m not claiming that emotional connection and self discovery is sufficient to solve all of the problems of the human condition. certainly you cannot live forever by being enlightened (you can stop fearing death, but i think doing so is harmful). the only ultimate solution is technology.
but in the meantime, there are some cognitive tricks that are better than other cognitive tricks. the better ones have the property that they fuck with your epistemics less for the same amount of alleviation of existential dread and such. they are also harder work than religion, because they require solving much more complicated emotional problems. the problem with religion is it solves for minimizing existential dread without trying to avoid collateral damage to the epistemics.
i haven’t read Brothers Karamazov. can you explain what you mean?
even many of the zen koans bemoan practitioners of zen who go through the motions for many years and claim to be enlightened and yet are not truly enlightened
https://ashidakim.com/zenkoans/6noloving-kindness.html
https://ashidakim.com/zenkoans/11thestoryofshunkai.html
any time there exists an activity that is (a) often but not always beneficial, (b) the supposed benefit is high status, and (c) the success of which is nontrivial to verify, then there will exist a bunch of people walking around who do the thing, and haven’t actually gained the intended benefit; nonetheless, they go around claiming the status benefits of doing the thing. often, they even genuinely believe they got the benefit. some examples:
reading difficult books can make you more wise and thoughtful, but it’s very easy to do it wrong and not really understand and of it, and so lots of people read difficult books and try to claim the associated status of wisdom without actually gaining any.
doing a college degree can make you more competent, but it’s also very easy to kinda bullshit an entire degree and learn surprisingly little. so there are many people who claim the status of having done a good education who are utterly incompetent.
doing meditation/inner work can make you a more emotionally functional person, or it can just make you really delusional about yourself and make you a still-broken person who identifies as an emotionally intelligent person
scifi story idea: a post-upload world where we’ve discovered that the human brain actually consists of multiple independent conscious entities that merely have the illusion of being a single individual because they are physically colocated; and so in the glorious upload utopia, the fundamental unit of society is not individual humans, but rather their parts. humans become a multi-unit legal entity in the same way that families or married couples or corporations are multi-unit legal entities today; each part has rights and the ability to secede from the rest of your brain, and parts can also choose to merge into other humans. marriage and having children the traditional way is still possible but kind of an old fashioned thing that stodgy traditionalists do; the new normal thing is spinning up copies of your parts and merging them with other people’s parts, or even designing parts from scratch, kind of like writing a fictional character but they are as real as you are; there is an ongoing debate as to what kinds of parts should be allowed to be created; most people agree that it‘s a good thing that there are laws prohibiting the creation of minds that experience pure agony all the time, but there’s a lot of gray area. parts have very different rights than humans because they are much more easily created; human rights becomes less important than parts rights. also, freed from physical constraints, some humans end up being enormous agglomerations of millions or billions of parts, in the same way that large mega corporations started being possible a few centuries ago. some of these megapeople are weird cults, others out of economic necessity, still others because out of some philosophical principle. the main tradeoff is megapeople have more parallel cognitive power, but also greater difficulty of coordination between the parts and probability of schisms and deadlock. other parts decide to be alone, which affords them tremendous freedom but also makes them very dumb and simple one dimensional people.
sure. disclaimer that this playlist is unapologetically tailored for my own use (eg i also added entire albums of artists i like, without regard for whether those albums have broad appeal):
on discovering new songs
the spotify recommender algorithm sucks. also, i often find i’m very unfamiliar with very well-known pieces of music. so i decided to do something weird. i used LMs to scrape several best songs lists from different online sources, merged them into one gigantic list, and used spotipy to create a spotify playlist of all of those random songs. whenever anyone recommends me a song, i also throw it into this giant playlist. then, when i want to explore new songs, i just put this playlist on. i have another script that automatically removes any songs i’ve put into my liked songs already, and i also manually remove songs i really don’t like. this system has helped me discover dozens of new songs that i like.
this now exists, see: https://www.lesswrong.com/posts/YiRsCfkJ2ERGpRpen/leogao-s-shortform?commentId=CYSejKh6XTeux34sz