Jesse Clark

@crancpiti @Acerola__t They can release code they created, though some of it they might have trains to want to keep secret (anti-cheat used in multiple games). They can't release third party services, which they're likely using. The server data shouldn't be made public for privacy reasons.

@Acerola__t The technical aspects of private servers. How difficult is it really for them to release the server files or source code? I think it should be easy to provide the server files etc as is and let the community figure it out how to run and maybe be required to help out

A very small number of people ruin it for everyone else: the problem with unmoderated social media sites in a nutshell. sciencedirect.com/science/articl…

“Surveys show that people greatly overestimate both firms' profit margins. For example, the SAEE found that the general public believed the average profit margin made by American corporations to be 46.7%, while the actual average that year was just 3%.” stevestewartwilliams.com/p/the-worst-ec…

@exfatloss @ESYudkowsky This study has the wrong control group for the hypothesis you're making. Also, the animals in groups are related, which would increase correlation between them. This increases the likelihood of a false positive. This is especially relevant given the tiny sample size.

Most depressing paper on linoleic acid ever: ncbi.nlm.nih.gov/pmc/articles/P… "To assess the contribution of dietary fatty acids, male and female mice fed a high-fat diet (35% energy as fat, linoleic acid:α-linolenic acid ratio of 28) were mated randomly and maintained after breeding on the same diet for successive generations. Offspring showed, over four generations, a gradual enhancement in fat mass due to combined hyperplasia and hypertrophy with no change in food intake."

@therealshanmao @ESYudkowsky Also after following the links in that source, I found them talking about a study in which mice fed almost no linoleic acid weighed less than those fed linoleic acid (the difference was compensated by saturated fat). This is best explained by linoleic acid deficiency.

@JFC_Bass_Chant @ESYudkowsky Ancestral levels of LA in fat tissue were around 2 %. Now it's at 10-20 % or even higher. So maybe we already have enough of it, and don't particularly have to worry about LA deficiency? fireinabottle.net/the-body-fat-s…

Say "linoleic-free", every time. "Low-linoleic". "Linoleic under 1% of calories". Otherwise you're going to get chicken fat and lard; non-ruminants pass linoleic in their feed straight through to you. Or you'll get non-tested "avocado oil". Or almonds that weren't pressed.

@therealshanmao @ESYudkowsky You don't have to worry about LA definitely on a normal diet. But it may still be possible with a diet that tries to avoid it.

@JJ_McCullough I said American because the scenes didn't seem very rainy to me. People were on sunny beaches, etc.

Do you think of the popular children’s book series “Where’s Waldo” as being American or British?

@ESYudkowsky They shouldn't serve indefinitely. They should be elected directly in a Condorcet system and a supermajority in the legislature should be needed to replace them. And there should be three chief executives who make decisions by majority and have staggered elections.

The worst common electoral system after First Past The Post - possibly even a worse one - is the parliamentary republic, with its absurd alliances and frequently falling governments. A possible amendment is to require 60% approval to replace a Chief Executive; who otherwise serves indefinitely, and appoints their own successor if no 60% majority can be scraped together. The parliament's main job would be legislation, not seizing the spoils of the executive branch of government on a regular basis. Anything like this ever been tried historically? (ChatGPT was incapable of understanding the question.)

@ESYudkowsky I don't think I've ever pronounced that vowel after a Y. Which I guess may be part of why it gets mispronounced.

P/S/A: The "Yud" in "Yudkowsky" is pronounced like the Hebrew letter "Yud" to rhyme with "would", "could", "should", or "good", rather than "cud" or "spud".

@Aella_Girl I'm pretty sure the average incandescent light is mostly infrared. And there are heat lamps used for raising chicks. So if someone's paying a lot for this, I'd guess they're being scammed. And I'd guess the infrared itself is a scam too.

health ppl sometimes are into infrared light for healing. they buy expensive light panels, use infrared saunas, etc. but like, isn't the sunlight a thing? like if ur not considering UV damage or whatever, can u not get the same benefits from sitting outside?

@DonaldH49964496 @ESYudkowsky If there's a carboxyl or hydroxyl group, the reaction can be broken down into simple reactions that aren't possible on saturated hydrocarbons. Hydride subtraction, proton donation and removal, resonance structures where oxygen shares the charge with carbon, etc. 2/2

@DonaldH49964496 @ESYudkowsky It's hard to see how. Most lower-energy reactions involve a single bond moving, and every atom having a full octet at each step. That would require carboanions in a saturated hydrocarbon. But those are fairly high energy. 1/2

The most reasonable guess by a true Outsider for which taste a biological organism would most enjoy, given the training cases for biology, would be "gasoline". Gasoline has very high chemical potential energy; and chemical energy is what biological organisms use, and what would correlate with reproductive success... right? If you'd never seen the actual results, "biological organisms will love the taste of gasoline" would sound totally reasonable to you, as a guess about the result of evolution. There's a sense in which it is among the most likely guesses. It's just that, on hard prediction problems, the most likely guess ends up still not being very likely. Actually, humans ended up enjoying ice cream. You say ice cream has got higher sugar, salt, and fat than anything found in the ancestral environment? You say an Outsider should've been able to call that new maximum, once the humans invented technology and started picking their optimal tastes from a wider set of options? Well, first of all, good luck to a true Outsider trying to guess in advance that the particular organic chemical classes of "sugars" and "fats" and the particular compound "sodium chloride", would end up being the exact chemicals that taste buds would detect. Sure, in retrospect, there's a sensible story about how it happened -- about how those ended up being used as common energy-storage molecules by common things humans ate. Could you predict that outcome in advance, without seeing the final results? Good luck with that. But more than that -- "honey and salt poured over bear fat" would actually have more sugar, salt, and fat than ice cream! "Honey and salt poured over bear fat" would also more closely resemble what was found in the ancestral environment. It's a more reasonable-sounding-in-advance guess for the ideal human meal than what actually happened! Things that more closely resemble ancestral foods would more highly concentrate the advance-prediction probability density for what humans would most enjoy eating! It's just that, on hard prediction problems, the most likely guess is still not very likely. Instead, the actual max-out stimulus for human taste buds (at 1920s tech levels) is "frozen ice cream". Not honey and salt poured over bear fat. Not even melted ice cream. Frozen ice cream specifically! In real life, there's just no reasonable way for a non-superintelligent Outsider to call a shot like that, in advance of seeing the results. The lesson being: Black-box optimization on an outer loss criterion and training set, produces internal preferences such that, when the agent later grows in capabilities and those internal preferences are optimized over a wider option space than in the training set, the relationship of the new maxima to the historical training cases and outer loss is complicated, illegible, and pragmatically impossible to predict in advance. Or in English: The outcomes that AIs most prefer later, when they are superintelligent, will not bear any straightforward resemblance to the cases you trained them on as babies. In the unlikely event that we were still alive after the ignition of an artificial superintelligence -- then in retrospect, we could look back and figure out some particular complicated relationship that ASI's internal preferences ended up bearing to the outer training cases. There will end up being a reasonable story, in retrospect, about how that particular outcome ended up being what the ASI most wanted. But not in any way you could realistically call in advance; and that means, not in any way the would-be owners of artificial gods could control in advance by crafty selection of training cases and an outer loss function. Or in English: You cannot usefully control what superintelligences end up wanting later by controlling their training data and loss functions as adolescents.

@Nunnie3001 @Lone_Swordsman1 @datepsych Generative AI doesn't save the training data, and it doesn't copy it to the output. They're designed to output things that have the same probability as the training data, but are still original. Different models do that in different ways.

@JFC_Bass_Chant @Lone_Swordsman1 @datepsych I'm not an expert (maybe you are )but I'm pretty sure Generative AI just takes all all the faces it ever seen in it an output that. there wont be an ugly person because the data took all the ugly to make a person ultimately not ugly

@Nunnie3001 @Lone_Swordsman1 @datepsych That's true, but AI doesn't take an average. It's trained to output the same distribution as the training data.

@JFC_Bass_Chant @Lone_Swordsman1 @datepsych If you take an avg composite of faces you're going to get someone attractive. I think it's a pretty well known phenomena (90% sure it's true)

@B1gMuddy @ESYudkowsky @Aella_Girl The wave amplitude in the other 60% of the space increased to compensate.

@ESYudkowsky @Aella_Girl Where did it go?

@Aella_Girl thanks for eliminating 40% of my possible Aella mass

@Lone_Swordsman1 @datepsych It's probably trained on as many images as the developers can find. Later reinforcement learning probably trained it to produce attractive faces. That's not surprising, the people evaluating probably preferred images with attractive faces.

@datepsych It seems that it's trained only on attractive people pics

@datepsych The humans doing RLHF must've given higher ratings for attractive faces, until the model only produced those. This isn't super surprising, but is funny.

@Aella_Girl We're currently at a stage where theoretical work can be done without a lot of technical knowledge, was long as it's specified in pure logic/mathematics. You don't need to be a programmer necessarily. Ontology identification is one such problem.

i feel generally ready and earnest but kinda incompetent. i feel like i'd be frodo at the council, short and hairy feet but like 'ok if u need someone to go walk into a volcano that seems pretty straightforward and i can do that'

i think i have a lot of learned helplessness around doing agentic projects associated with worldsaving stuff. i feel kinda dumb and not technically minded and i just auto do some submission 'im bowing out i cant think about this not my field' mental move

@ESYudkowsky It would technically be possible for a single double bond to be enough to make it possible for organisms that evolved for it to break down a hydrocarbon. Though it might have the problem of being less water soluble, and more likely to evaluate. 2/2

@ESYudkowsky I don't know of any low energy reactions that can break down pure saturated hydrocarbons, except those that use highly reactive compounds that would likely react with everything else in the organism. Fats can be broken down much more easily. 1/2