prerat

@tszzl you have misunderestimated chalamet. only the star of wonka (2023) could be the voice from the outer world

the dune movies were doomed from the start to be good and not great due to the casting of chalamet as paul. he does not have the gravitas for a child-god and is much better suited for kind of silly coming of age movies

@shlevy i think this gets to the core limitation of current ai. within the context window it's pretty sharp at picking stuff up, not superhuman but like a human thinking hard for 15 min but theres no equiv to human learning something for a few days/weeks/months. smth sample efficiency

No, but any human who could write code like LLMs can could learn brainfuck and write something passable in a few weeks. Or just write a compiler targeting brainfuck.

This doesn’t prove it’s a bubble, but if these systems were in fact intelligent in the sense and to the degree that many proponents claim, they would obviously be able to write decent brainfuck without having seen it before.

@MasterTimBlais @mechanical_monk congrats tim but also accept my old waiting chesscom friend request so we can chess

2.5 years ago, @mechanical_monk gave me a single chess lesson. this lesson would grow to define me. it would consume countless hours of my life and possibly destroy what remained of my ability to work on anything ever today i finally crossed 1500 elo on chesscom. thanks monk

@ChanaMessinger bad guy says "EVERYBODY ON THE FLOOR"

What exactly is the disaster where crouching on the floor would really help you?

man-spider plot in the new spider man movie???!?!? its risky bc appealing to 90s kids has historically gone badly, e.g. raimi spider man 3 with "aggressive" symbiote (originally from 90s spider man tas show) but otoh i am the 90s kid so i will show up for it

@nosilverv idk doesn't it feel bad to win just because your opponent had bad luck, instead of thru your own skill? in my head glhf is saying like let's have a good competitive match where the better player wins

There's a weird thing I don't really understand about ego… Like in the case of 'good sportsmanship'. In magic online people often write 'glhf' i.e. "good luck have fun". I feel torn—I want to say "good luck" back but I don't actually wish for them to have good luck: I want to win. Similarly I don't believe they actually want me to have good luck: they want to win. But it feels like we "should" (""""""should""""""?) not be attached to the outcome. Like that is somehow base. Below us. But we're also definitely playing to win?? Or like, "argument" and "rationality" in politics. People—from both aisles, mostly—pay lip-service to this. But how many of those people can even say what is a premise or what is soundness or validity? In both cases—in games and in politics—identity and the desire-to-win/crush-your-opponent seems like HYUUUUUGE motivators but that somehow cannot be acknowledged? And so there's this weird co-conspiracy of mutual bad faith where I won't call out yours if you won't call out mine… But in fact people DO NOT know what makes a good argument and they DO NOT overcome their (supposedly) base/selfish desire to win. Which opens the question, to me: would it be better if we were like? "Fuck you have bad luck I hope you die"? Or "fuck arguments IGTFKY"? Like I just don't see a path from the current thing—the simulacrum of the good thing—to the desired thing—the actual good thing…

im not surprised that meta is killing off horizon worlds on quest. it was like sorta worse homogenized vrchat. i AM surprised that meta is KEEPING horizon worlds on mobile. huh?? am i out of the loop and kids are playing that instead of roblox?

reddit and twitter feeds are fundamentally different (i think) subreddit feeds are (roughly) globally ranked, and then your feed is roughly just the subreddit feeds combined twitter for you is individually sorted (roughly) on your individual P(like). NOT global like count!!

a dislike button could work but you have to be smart about it with the algorithm for example, if i have someone blocked, their downvotes should count as upvotes for my feed (and their likes should count as downvotes)

@AustinA_Way @openclaw where's this chart from? it seems to say chatgpt has had no impact on cheating bc the overall cheating rate has stayed the same

There's a cheating method so advanced that no school in the world can detect it. We just caught someone using it. An AI agent can now complete online coursework indistinguishably from a real student. Tools like @Openclaw can mimic the arc movement of the mouse, the pace of a student, etc. It operates at the system level, meaning there's nothing for a platform to detect. So how did we catch someone using it? Every student leaves a pattern in their data, and when an AI agent does the work, that pattern looks nothing like a real student. By extrapolating student data and comparing to the mean, you catch a cheater.

@DeepDishEnjoyer the p is for pat modesto

me to yudkowsky: "suppose you were wrong in your assessment about your status as a great thinker" yudkowsky: "but i am a great thinker"

@nosilverv @Alphiloscorp it might be exhaustive but "cause" is maybe doing outsized job -- any question of "why" could be phrased as confusion about cause, and once you understand "why" then you aren't confused anymore

Has someone made a taxonomy of confusion? Is this exhaustive?: • What is there? → count / identity / parts-wholes •Where are the boundaries? → categories / cuts / levels •How is it related? → cause / dependence / structure •What does this term pick out? → semantics / reference •What kind of status does it have? → descriptive / epistemic / normative / modal / temporal

@RyanPGreenblatt to me it kinda seems like so much of an understatement that it's almost dishonest in the other direction "We're racing to superintelligence because we want to cure death" would be more true but also would probably freak ppl out

"We're racing to superintelligence *because* we want to cure cancer" may be a lie, but ASI would likely be capable of quickly curing cancer. I think the risk are very high and well timed slowdowns are easily worth it, but this doesn't mean the benefits of safe ASI are small.

@ShadowyZephyr well it's at least directionally useful

Am I the only one who still thinks the term "directionally correct" is useful

@tenobrus fwew its not me, recent time mutual 🤓

is it okay to unfollow a very very long time mutual if i've slowly realized over the course of years that they have incredibly unfathomably retarded takes on every possible subject and to make up for it are also unpleasant and abrasive

@abrakjamson but is your self worth based on being CDT rational or FDT rational

@prerat You can't convince me, because no matter how outlandish the scenario makes my opinion seem, I derive immense self-worth by being more rational than everyone else!

in the year 2067, you are approached by an all powerful acausal Mr Beast he says he secretly flipped a $10M coin a minute ago he then explains to you the rules...

@tenobrus @mimi10v3 i voted bsky because most niche. the argument was strongest when it was mostly just tpot on bsky when i made this post, but i think it probably still applies somewhat

@mimi10v3 people are voting bluesky because of the deranged leftists but they're forgetting twitter is still the prime location for edtwt / shtwt / mentally ill teenagers in literal cults / etc

which platform is the most mentally ill?

@staysaasy compromise w half birthday party with minor present ? do not claim it is full birthday but still allows some fun and introduces the concept of 0.5

2.5yo is convinced that next week is her birthday. Nobody has told her that it's her birthday, she can't read a calendar, and doesn't understand time very well. How do I deescalate. She is making a list of presents that she wants.

@moonstne the tricky part is that you didn't agree ahead of time when he tells you the rules, you then are probably like "dang i wish i had precommitted to pay" but you haven't & don't have a chance to before he reveals the coin flip result, which is negative so can you pay anyway?

@prerat If I agreed to the terms before the flip, and the coin was 50/50 odds...or odds that I agreed to, then sure, I would pay the $6.7

@RokoMijic yeah this is "counterfactual mugging" from LW (w mr beast flavor text) i just think it's interesting bc it tests your conviction more, bc you're losing in this branch (vs winning in newcomb). & it doesn't have the moral intuition that sometimes makes parfit's hitchiker easier

@prerat This is just Newcomb's Problem... if you would one box on Newcomb you should give him $6.7 Of course that is assuming you believe the setup, otherwise it looks like a Pascal Mugging