time to plurb out
26.11.2025 21:11 β π 1 π 0 π¬ 0 π 0@iid.bsky.social
A collection of three IID random variables stacked on top of each other in a trench-coat she/her
time to plurb out
26.11.2025 21:11 β π 1 π 0 π¬ 0 π 0"oh its meaning indicated by the presence of ne, which is commonly dropped. have fun"
26.11.2025 20:23 β π 2 π 0 π¬ 0 π 0"jamais" drives me up a fucking wall
26.11.2025 20:21 β π 2 π 0 π¬ 1 π 0Crusts???????
26.11.2025 17:49 β π 0 π 0 π¬ 0 π 0I really wonder what the check out guy who definitely knows me thinks of me racing into the shop 10 minutes before close to buy three pie pastries
26.11.2025 17:49 β π 1 π 0 π¬ 2 π 0Just bought 20 euros worth of different types of pumpkin purΓ©es in the hope one will work for pie
26.11.2025 17:36 β π 0 π 0 π¬ 0 π 1Cranberry sauce is uncomplicatedly good
26.11.2025 16:48 β π 2 π 0 π¬ 1 π 0they already look like a toxic corporate lawyer x PR manager power couple with a brownstone on the east side and a completely unmanageable four year old who keeps biting other kids at his Montesori school, let's make this real
26.11.2025 15:01 β π 1 π 0 π¬ 0 π 0I think Ariana Grande and Cynthia Erivo should get married for real. I think everything that would involve and devolve into would be incredibly funny and would keep too online lesbians occupied for deacdes
26.11.2025 14:57 β π 5 π 0 π¬ 2 π 0it's unfortunately very real though for the fate of the land to turn on the toxic relationships spawned by one freshman friend group from an Ivy League university, the majority of which immediately became evil upon graduation
26.11.2025 14:46 β π 3 π 0 π¬ 0 π 0still thinking about Wicked: it did feel very timely. Wicked is variously cynical in every adaption I've ever seen. the musical typically less so, the book extremely so. the movie leaned more towards the book side of the spectrum, the ending is bleak about what's necessary to effect political change
26.11.2025 14:44 β π 0 π 0 π¬ 0 π 1I think thereβs definitely a way to combine stochastic string generation and concept mapping with a strict logical model that could make theorem provers more βcreativeβ in a sense, but only when used in tandem
26.11.2025 13:53 β π 2 π 0 π¬ 1 π 0I agree that *maybe* you can extract something like a logical structure for the strict generation of language, in the same way that an FSA can be used as a logical descriptor of the string set it generates
but that's about existence, not about application to a utility
"concept map" that maps words and syntax to a metric space with a meaningful distance measure---which i think is true-ish---that still does not imply a jump to the existence of internal logic. how could it? that's a completely different mathematical structure
26.11.2025 13:45 β π 1 π 0 π¬ 1 π 0the continued insistence from essentially everyone but mathematicians that there simply must be some kind of logic engine hiding in there, some kind of human conceptualization of a chain of thought predicated on an algorithmic understanding of the world drives me crazy, even if there's a fuzzy...
26.11.2025 13:45 β π 5 π 0 π¬ 1 π 0god the level of illiteracy around this tech is *unbelievable*
you can't tell! you will never be able to tell! that is a mathematically unanswerable question in any way other than just explicitly describing the weights that go into a probability mixture that's then sampled
we know this because Anthropic acted incredibly shifty during the entire libgen scraping case and has told multiple deliberate lies about data procurement. *they may not even know what's in there.*
26.11.2025 13:38 β π 1 π 0 π¬ 1 π 0there is no inverse function from output to data for an LLM. you cannot trace why an LLMs recommends a product back to the data that went into the soup, and I strongly suspect that questions about what data went where are literally unanswerable---buried on hard drives that have since been wiped
26.11.2025 13:37 β π 4 π 1 π¬ 1 π 0this is so naive it almost makes me cry. that first L in LLMs specifically refers to the total inability to do this. the jump in language models occurred when OpenAI threw out all standards for data scraping---something ML used to be very concerned about---and took a massive kitchen sink approach
26.11.2025 13:36 β π 8 π 1 π¬ 1 π 0I wouldn't be surprised if we actually see internet use start to drop as social media gets less interesting and all easily accessible content becomes a sea of samey AI slop. the weirdos--everyone on here, for instance--will keep pursuing places to be weird together. but normies? why bother?
26.11.2025 13:30 β π 5 π 0 π¬ 0 π 0my prediction for the internet is that original, interesting content will be increasingly restricted to walled gardens: self-hosted forums, fan spaces, discord servers, maybe there's space for some independent social media like bluesky, maybe not. the rest of the internet is going to be smoothed out
26.11.2025 13:28 β π 5 π 0 π¬ 2 π 0I've been saying this for literally three years now and it's crazy how few people seem to have drawn this very direct line. LLMs fundamentally can't work as content aggregators because they by their nature suppress genuine content creation
asimovaddendum.substack.com/p/are-llms-t...
This itself is pretty easy to disprove---anyone who spends time with a child knows that the failure modes of a human brain in development are very different than the failure modes of an LLM---but it's disturbingly persuasive to a lot of people, especially those who seem to be broadly alienated
26.11.2025 13:15 β π 6 π 0 π¬ 0 π 0Anyone making this argument knows very well that the mathematics of GPTs are not so esoteric as the boosters pretend, and in fact are very easy to describe, and so absent some appeal to a vaguely-defined "emergence" the only argument can be that the mathematics ARE how human brains work.
26.11.2025 13:14 β π 7 π 0 π¬ 1 π 0The thing is this isn't about believing LLMs have become humans, but rather the belief that humans are essentially LLMs. The viewpoint driving this is that a human brain is just moving around probability distributions on strings, arguing about LLM consciousness is a distraction
26.11.2025 13:12 β π 15 π 5 π¬ 1 π 0i don't know, he got want he wanted
26.11.2025 13:08 β π 10 π 0 π¬ 2 π 0I don't think betting "Russia will not escalate" is a super smart bet in this decade. Just personally.
26.11.2025 13:06 β π 5 π 0 π¬ 1 π 0Listen like, I don't like to opine on what Russia will/will not do because frankly I don't think I have any sense of it, but I do think I have the sense to call out Americans who are acting like "Russia cutting undersea cables" is a total impossibility when they do this very often in the Baltics
26.11.2025 13:05 β π 5 π 0 π¬ 1 π 0I do like the idea of going the other way so characters have to adopt their actors sexuality and suddenly most classic American tv characters being bi
26.11.2025 12:57 β π 1 π 0 π¬ 0 π 0Thereβs so much American stuff I just sincerely really like in an uncomplicated way. But also itβs an insular place with very high cultural walls
26.11.2025 12:56 β π 2 π 0 π¬ 0 π 0