nim

Karma: 1,220

nim Mar 31, 2025, 2:59 PM
2 points
0
in reply to: KFinn’s comment on: Policy for LLM Writing on LessWrong
Maybe. I think there’s a level on which we ultimately demand that AI’s perception of values to be handled through a human lens. If you zoom out too far from the human perspective, things start getting really weird. For instance, if you try to reason for the betterment of all life in a truly species-agnostic way, you start getting highly plausible arguments for leaving bacterial or fungal infections untreated, as the human host is only one organism but the pathogens number in the millions of individuals.(yes, this is slippery slope shaped, but special-casing animal welfare seems as arbitrary as special-casing human welfare)

Anyways, the AI’s idea of what humans are is based heavily on snapshots of the recent internet, and that’s bursting with examples of humans desiring animal welfare. So if a model trained on that understanding of humanity’s goals attempts to reason about whether it’s good to help animals, it’d better conclude that humans will probably benefit from animal welfare improvements, or something has gone horribly wrong. Do you think it’s realistically plausible for humanity to develop into a species which we recognize as still human, but no individual prefers happy cute animals over sad ones? I don’t.

nim Mar 27, 2025, 4:24 PM
2 points
0
in reply to: Alene’s comment on: Policy for LLM Writing on LessWrong
“Humanity” is a weird word at the moment. I think it’s more of a “descendants of our ancestors” thing—I think LLMs trained on humanity’s content should probably be viewed as related to us in important ways, ways that a hypothetical LLM trained only on interaction with octopi or slime molds would not be as related. But this may be a weird view, so let’s ignore it henceforth.

I think the “benefit humanity” rule is actually much broader than you’re reading it as:

Secondly, as an animal advocate, I want to preserve the opportunity for AI to make a post that will benefit animal welfare, even if the post doesn’t benefit humanity.

Your welfare is intertwined with that of animals. You are distressed by their suffering. Therefore, improvements in animal welfare would be expected to cause improvements in your welfare. I think an AI making a post which benefits animal welfare would benefit humanity, because it would be good for all the humans who will feel better in a world where animals suffer less. To put it simply, I claim that all posts which benefit animal welfare in a way that’s legible to you are also benefiting you. Kind of a big claim, but I can’t come up with a counterexample—maybe you can?

Since there are humans who care about AI wellbeing and are upset by the possibility that AIs could be suffering needlessly, it seems to follow that a post which reduced preventable suffering for AIs would benefit those humans.

The rule isn’t demanding that posts benefit ALL of humanity. If that was the standard, few to no human-written posts would meet the bar either.

nim Mar 24, 2025, 4:50 PM
4 points
0
on: Delicious Boy Slop—Boring Diet, Effortless Weightloss
Thank you for sharing this! Does it scale well with varying levels of physical activity for you? Some dietary things tend to be great at higher activity levels and fall apart at lower ones, and vice versa. I’m curious roughly what your current activity level is looking like and whether the boy slop diet thing has worked well at higher/lower levels as well.

Since this post seems a potential gathering place for other extremely easy and nutritionally adequate recipes, I’ll also leave my boring-food go-tos here—these aren’t as cheap as yours but they’re similarly easy:
- Bachelor bowls: can of black beans, rinsed. Can of sweet corn. Optional, tiny can of olives. Optional, thaw some frozen hummus or guac (costco sells single serves fresh, both of which freeze great). Mix, heat if desired, add salsa (canned is fine). I find that 2 cans makes 2 16oz mason jars of food, and one of those jars is a meal for me. A jar of this is my go-to when I have to pack a lunch.
- Oatmeal. Put a bit of butter and a cup or so of steel cut oats in a pot, toast them till they smell nice. Add some salt at some point—start with about a half teaspoon per cup of oats if you like measuring things. Add 3-4 cups water per cup of oats (I add boiling water, because it’s easy to make in the electric kettle and parallelizes the “heat water” and “toast oats” steps, but you can add cold water and let it take longer). Optionally, add some amount of dried fruit. Simmer, stirring occasionally, for 20-30mins until you like the texture. Refrigerate it right in the pot. When hungry, scoop some into a bowl, add a bit of milk or water, microwave till hot. This is my favorite dish to eat as a brain reset for a couple days when I lose track of the difference between feeling hungry and feeling bored.
- The instant pot thing: Put whatever vegetables like being well cooked (potatoes, carrots, etc) into instant pot. Add whatever cut of lean meat was on sale. Dump in a can of enchilada sauce. Cook on the meat cycle. Canned enchilada sauce gets the seasoning right and still tastes good after being exposed to high heats. I’ve tried it with other canned sauces and some of them are ruined by the heat.
One aspect common to all these easy foods is that they don’t require much management of perishables. Your tomatoes are canned; the beef can be frozen; the veggies should be frozen. I think people who talk about healthy eating tend to underestimate the difficulty that some people have managing perishables, especially when cooking for 1, so I appreciate that you’ve tacitly circumvented the whole problem in your advice :)

nim Mar 24, 2025, 3:47 PM
4 points
0
on: Solving willpower seems easier than solving aging
“things that I expect I will be glad I did once I’ve done them”

I like this definition because it allows 2 different routes toward improvement, which you’ll probably need to mix to get the best results:
1. Increase time spent on the activities which are already on your “productive things” list. Taking this to the extreme would likely eliminate load-bearing forms of rest, and drastically increase burnout risks.
2. Improve recovery activities to bring them from “unproductive” to “productive”. You hint at this with reading a good book making the “productive” list but reading a disappointing book falling into “unproductive”.
The lowest-hanging fruit for me tend to be in (2): When I’m able to give myself an alternative activity that’s equally easy but less regrettable compared to an “unproductive” habit, and notice when to do it, I can often get equivalent results for the function that the bad habit was serving.

nim Mar 19, 2025, 8:45 PM
2 points
0
in reply to: Said Achmiz’s comment on: Boots theory and Sybil Ramkin

It certainly is not true on Earth.

Where is it cheaper per item to buy the same food or household good singly rather than in bulk?

Where can you get a single roll of the same toilet paper for less per roll than getting it in a bigger pack? Ok, now what if you need TP and the cash at your disposal at that moment is less than the cost of the bulk pack but more than the cost of the single roll?

When is it cheaper (time + money) to cook one meal at a time versus meal prep for the whole week? OK, now what if you can’t afford the whole week’s worth of food at once?

We might be talking about poverty at different orders of magnitude, and you might be writing off a lot of failures to purchase efficiently as “skill issue”… but being poor in skills and the capacity to hone them is, itself, a form of poverty.

nim Mar 18, 2025, 11:14 PM
13 points
9
on: Boots theory and Sybil Ramkin
Invert your thesis, from “the rich are so rich because they spend less money” to “the poor are so poor because they spend more money”, and it becomes much more defensible. Boots theory captures the “being poor is expensive” element that’s true in Ankh-Morkpork and also true on Earth—look at the markups on any household item when it’s sold individually rather than in bulk, and then consider what happens to people who don’t have what it takes to plan ahead and make bulk purchases.

If we’re being particularly literal about things, Ankh-Morpork canonically lacked fiat currency until the events of Making Money, so we might want to consider the other ways in which their economics functions differently from our own. They also have magic, and the deities who implement their laws of nature sometimes pop round for a cup of tea. So I wouldn’t rule out that Vimes’s use of boots theory as a full explanation for economics could be less descriptive than prescriptive—he lives in a world where enough people believing anything hard enough makes it real.

nim Mar 10, 2025, 5:49 PM
8 points
3
on: when will LLMs become human-level bloggers?
IMO, excellent blogging happens when people notice connections between ideas while experiencing the world. Good blogging often feels like a side effect of the author’s learning process. The author was there when they formed an idea, there when they had experiences with the concept of the idea, there when they revised the idea… so they can report on the experience of making a lasting change to their understanding. Or if not directly reporting on the experience, they can at least generate novel entropy which increases a human reader’s odds of following the author to a particular new or interesting concept which might not be directly transmissible through language.

LLMs appear to lack the kind of “background” thought process that causes one to notice connections one isn’t explicitly seeking. They aren’t really “there” during the processes in which they assimilate information, at least not in the way that I’m still myself while reading a textbook.

I think if we wanted to get blogs out of LLMs that are good in the ways that human blogs are good, the most promising approach would be to try generating blogs as a side effect of their training processes rather than by prompting the finished assistants. But that touches the highly secretive stuff that the stewards of the most powerful models are least inclined to share with the world.

To step back a bit, though, do we need LLMs to be good at blogging? If we model blogs as demonstrations or tutorials of ways you can use a human brain, it doesn’t seem at all obvious that they’d keep their benefits when generated without a human in the loop.

nim Mar 6, 2025, 7:22 PM
7 points
5
on: The Hidden Cost of Our Lies to AI
Don’t people notice the ways that seeing themselves engage in particular behaviors updates their own self-image? The payoff for being polite to language users is that it makes me see myself as the kind of person who is generally polite. The results of being mean and bullying would be that I would come to see myself as the kind of person who is okay with engaging in mean and bullying behavior to get what they want.

Or maybe everybody else has a skill at compartmentalizing which I lack? But I absolutely catch myself applying prompting strategies to human conversation, because the part of me which does that self-concept feedback loop doesn’t differentiate between my behaviors toward animate audiences versus my behaviors toward “inanimate” ones.

nim Mar 4, 2025, 4:50 PM
4 points
1
on: The Semi-Rational Militar Firefighter
Yes please, more please!

Your writing is more enjoyable than that of many native English speakers, and I am one.

Will you do a sequence with more stories?

Tell them the way that’s fun to tell, like here, and they leak rationality. The “systematized winning” kind, because it’s all rooted in finding ways not to die in situations where statistics would suggest you really ought to.

nim Feb 24, 2025, 10:59 PM
2 points
1
on: The manifest manifesto

Think, at least a little, before you act in a new or complex way.

Oh. Is this what they meant by “think before you act”? Because thinking before every action goes exactly as well as you, careful author of that sentence, would expect.

There are at most as many gods as people.

why can’t we all have unique pantheons?

nim Feb 13, 2025, 10:50 PM
4 points
0
on: I’m making a ttrpg about life in an intentional community during the last year before the Singularity
Intentional symbolism to place the singularity, start of a whole new everything, in the springtime of the second year?

I would try playing it. Will be interesting to see how it plays at a mostly non-rationalist table :)

nim Feb 13, 2025, 6:43 PM
5 points
0
on: Studies of Human Error Rate
We put decades of work into getting software to behave less like databases, and then act surprised when it doesn’t behave like a database. C’est la vie.

nim Feb 12, 2025, 11:30 PM
4 points
0
in reply to: Jonas Hallgren’s comment on: Why you maybe should lift weights, and How to.
what set of exercises do you prefer to strengthen and stabilize the rotator cuffs?

nim Feb 12, 2025, 11:25 PM
6 points
0
in reply to: Richard_Kennaway’s comment on: Why you maybe should lift weights, and How to.
I’m not near any gyms and have plenty of space, so I strongly prefer the safety profile of barbell + squat rack. If something goes wrong and I lose control of the weight, it’s good to know that there’s no way for it to hit me. And part of progressing is sometimes trying a weight that you’re only 80% or 90% sure you can actually lift successfully—I’d much rather the failure mode be clank “whoops!” than however many pounds of iron to the face.

I also find that it’s easier to track whether I’m using good form with a barbell vs dumbbells. The cues to moving a bar correctly are to keep it level and trace the correct path in a plane with it (usually straight up and down), whereas the cues to moving dumbbells correctly are… something about simultaneously tracking the position and orientation of two separate objects in 3-space? Definitely more complicated to keep track of all the moving parts when there are twice as many.

Barbells are also a compromise between fixed and adjustable dumbbells. You’ll probably want different weights for different exercises, and the exact weight you want for each will slowly increase over time. With a barbell, you modify the two ends to change the weight, and you can get plates as small as a few ounces if you need to increase the weight super slowly. With dumbbells, either you use fixed increments and have to store a lot of pairs, or you use adjustable ones and have to modify twice as many ends for each weight change.

If your access to equipment, space, and trainers is different from mine, though, your conclusions about the best options for training will likely be different as well.

nim Feb 7, 2025, 6:21 PM
2 points
0
on: Wild Animal Suffering Is The Worst Thing In The World
eliminate predation.
Ok, I’ll bite—who was doing the predation, and what are you suggesting ought to happen to those creatures?
I agree that adding preventable new suffering is bad, but I don’t follow that into any obviousness that it’s good to meddle deeply in nature’s feedback loops. To oversimplify, let’s imagine a button that releases a virus which painlessly inflicts all living beings with https://en.wikipedia.org/wiki/Congenital_insensitivity_to_pain. I think it would be bad to push that button, but the way you’re describing suffering creates a mental model for me in which you would push it. I’ve probably missed something about your perspective that makes it seem obvious to you why that wouldn’t be the case, but I hope the example highlights which important bit turns out to need to be said out loud to get the idea across to others.

nim Feb 7, 2025, 5:37 PM
13 points
3
on: How AI Takeover Might Happen in 2 Years
guerrilla, not gorilla, warfare :)

a good read. thank you.

nim Jan 31, 2025, 6:46 PM
3 points
0
on: 5,000 calories of peanut butter every week for 3 years straight
have you ruled out the powdered peanut butter options? They tend to come in at under half the calories of regular PB, while getting pretty close in replicating the flavor and texture. Assuming you’re serious about the psychological impact of removing all peanut butter products from your diet, at least.

nim Jan 29, 2025, 12:22 AM
5 points
0
on: A Three-Layer Model of LLM Psychology
Feels to me like at the moment, the “character layer” handles transforming a “you” into an “I”. I find it conspicuously absent in smaller models running locally, though maybe the absence is just more obvious with CoT than without it.

I’ve also noticed that the “training data” we get as humans is primarily focusing on or contextualized relative to ourselves, whereas the parts you’re referring to as ground layers don’t really have a concept of the LLM as an entity so they tend to assume they’re humans on that level.

nim Jan 24, 2025, 9:45 PM
2 points
0
on: nim’s Shortform
saw a post from zvi on twitter yesterday, “remember that this corner of the world has a very different outlook than most other corners. if you don’t have random day-to-day ai queries what are you even doing all day?”, in reply to someone asking “who the fuck has random day-to-day ai queries?” (yes, i looked at that on my phone and typed it out, because my twitter account for following tpot people lives only on my phone, and my lw account lives only on a device with a real keyboard, and getting the link from point A to point B would have been even more hassle)

i think the lines between worlds could be defined relative to search. like, in the beginning, there was no search. if you needed information, it was given to you. if it wasn’t given to you, you made do without it. some people still live there.

then there was search, and some of us learned how to use it (others definitely didn’t!) i’ve had to train my parents carefully to search, and my mother has taken to it better than my father. search at its best, with grep within a single local text file, is amnesiac, immaterial and eternal—there is no social calculus around having searched becuase the social implications of “over there on the computer” are unrelated to the contents of the typing and reading going on. i think I speak search with a faint accent, like a kid who moved countries in their early teens.

AI feels to me like it’s at its best when I use it as search on ideas. It also works as a bigger room to layout concept prototypes in, but that’s a higher barrier to entry; it’s harder for me to learn to design and build bigger conceptual things than it is to just use good search as search.

so we have two search modalities, conventional/literal and linguistic. perhaps worth noting, i don’t think modern google is the embodiment of the old-school literal form of search—i think database queries are what that grew up into as it got with the times. Modern google and google-alikes feels like it’s a database trying to be the kind of concept-search that LLMs surpass it at.

it feels to me like the question of which search tech is superior is answered by the context of what you’re searching for. it’s the difference between numbers and letters, the difference between data and information, the difference between intelligence and wisdom—it’s a level-of-abstraction kind of distinction.

thus the question becomes, for those of us who do queries all day, what are we querying? DB queries are for things with right answers, because they can kinda read a smudged copy of the answer key to the test of reality. AI queries are worse at that, but they’re better at a different kind of thing.

There’s also a kind of creativity and engineering where you query reality directly. There are areas where direct-query is a lot better, and other areas where db-query is a lot better… like if i wanna know about how this one item will behave, i should examine it, whereas if i want to know how items of this type should be expected to behave, i should not extrapolate over-much from my examination of a single item. it’s like something not-turning-up-in-search that i recall seeing recently, perhaps a youtube short from chris boden, about the difference between “engineering” and “the knack”—if you have the knack you can overbuild anything, but engineering is the art of cutting away all the excess to build the bare minimum-viable of any given thing. If you’re using the knack, it gets to the point where you ask the materials what they’re gonna do, and they give you a better answer than the books about them might.

but why would people close with AI have that physical experience? Material hobbies are expensive and messy and cost a lot of storage space and waste time you could spend arguing about x-risk, and when material hobbies do arise, computer-people have the means to get idealized standardized raw materials for them from wherever in the world they’re made, rather than having to make do with whatever you can find nearby. Yes, there are exceptions. Yes, you’re exceptional. But most people? Have you ever tried teaching a room full of most-people to tie a new knot? Modern life very rarely calls “us people” to direct physical problem-solving, us who do so well or at least so connectedly online. So of course we’d forget the other way, the get-your-information-straight-from-the-source way, if we veer away from the areas where that way’s better.

There are probably other hobbies where this is relevant. Maybe you like hiking, maybe you sail. You are probably worse at predicting the weather than the forecast on your phone, but you’ve probably met a human who was better. It’s kinda like that, with the kind of queries that lose resolution if you even put them into language. Most people building something real have to go through a quantified level, a CoT corrigibility level, to be adequately assessed and supervised on their thought process. At least, that’s the engineering-degree paradigm. And the engineering-interview one, or at least the fields that larp as engineering, “solve this problem on the whiteboard and show your work”. It’s an implicit prioritization of the kind of work that can be shown.

anyways, that’s all to say that i think there’s a certain kind of smart that’s agnostic of whether you’re expanding your worldview in the externally legible way or the less-legible ones. i think there’s a “kind of guy” who is basically the same “kind”, in the meaningful ways, as the constantly-querying-AI one… but the kind who queries the world instead of the world-model. It probably takes both kinds. What’s “it”? Who knows!

nim Jan 22, 2025, 3:53 PM
2 points
0
in reply to: Viliam’s comment on: Don’t Legalize Drugs
I have a lot of elderly neighbors. Alcoholism’s been a major factor in about half the deaths in the past few years. The runners-up in this popularity contest are cannabis of the grow-it-on-the-sly strength, and opioids because they’re nasty but people tend to start them later in life.

If we’re being literal about drugs as “non-caloric chemicals we ingest to modify our physiology”, the most popular drug in the elderly is probably a blood pressure med of some kind.