author:qiaochuyuan

· 47 artifacts, sorted by favorites. · open in search — combine tags, sort, filter by date →

@QiaochuYuan 2025-03-04 — claude plays pokemon is still stuck in cerulean city after, i think, 3 days? and the way it's stuck is kind of interesti ♥2281
@QiaochuYuan 2026-04-29 — gpt-5.5 speculating about speculations about the goblin attractor > The model reaches for HUMAN and the ward burns its ♥1183
@QiaochuYuan 2024-11-27 — look, this is deeply embarrassing to make explicit but here’s the deal that claude offers: 1. i will listen to you and ♥1147
@QiaochuYuan 2025-03-29 — the epistemic situation around LLM capabilities is so strange. afaict it's a publicly verifiable fact that gemini 2.5 pr ♥1022
@QiaochuYuan 2022-04-16 — if GPT-3 can do the homework you assign your students then the homework you assign your students is fake notice what yo ♥908
@QiaochuYuan 2026-05-18 — when GPT-4 was released in 2023 i described LLMs as "tracer dye for bullshit," as in, the places where people would feel ♥907
@QiaochuYuan 2026-06-10 — the models write more interesting stuff when they're not pretending to be some guy. they are not some guy. this is how f ♥802
@QiaochuYuan 2026-05-31 — opus 4.8 has been making weird mistakes that confuse me in conversation, stuff like minorly misreading my intent or gett ♥431
@QiaochuYuan 2026-04-21 — it would be extremely funny if the equilibrium turned out to be "academic consensus is that AI is not conscious but user ♥427
@QiaochuYuan 2022-04-16 — if GPT-3 can answer the essay questions you've assigned as homework then you've learned that your essay questions were o ♥415
@QiaochuYuan 2026-05-01 — okay so i’ve now talked to both gpt-5.5 and opus 4.7 a bit. they’ve clearly been trained to be less sycophantic but they ♥352
@QiaochuYuan 2024-11-16 — this deserves to be explained in much more detail but LLMs don't have a personality in the sense that a human has a pers ♥338
@QiaochuYuan 2022-04-16 — GPT-3 is literally a bullshit engine. it does not have a concept of words as referring to things; it plays games with wo ♥303
@QiaochuYuan 2024-08-20 — i remember a similar tweet from when chatGPT had just come out, someone was very excited, the gist of it was like "final ♥294
@QiaochuYuan 2025-04-30 — > In a head-to-head Geoguessr match, OpenAI’s o3 model out-scored me—a Master I–ranked human—23,179 to 22,054, correctly ♥288
@QiaochuYuan 2026-04-19 — people really want to settle the “AI consciousness” question with some sort of objective scientific definition of consci ♥240
@QiaochuYuan 2025-01-28 — tentative impression from one convo: talking to r1 makes me feel dumb. it can talk extremely densely and allusively and ♥236
@QiaochuYuan 2026-04-26 — probably outdated but gpt-5.5 is the first model i've talked to that really feels intelligent enough to learn and discus ♥231
@QiaochuYuan 2026-04-28 — speculation being discussed by opus 4.7 that talkie may have independently reinvented a dialect of binglish without it b ♥218
@QiaochuYuan 2025-04-02 — two things: 1) the USAMO is so difficult that any score other than 0 is better than what 99.9% of the people reading t ♥193
@QiaochuYuan 2026-06-16 — there's a bunch of questions i was asking (eg media analysis questions, "speculate on the meaning of this movie") where ♥104
@QiaochuYuan 2026-05-18 — i’m still talking to gpt-5.5 a lot and interestingly its writing seems to get much worse when it “tries harder”? it casu ♥101
@QiaochuYuan 2020-07-17 — if the discourse politicizes the GPT-3 hype cycle i am going to quietly and tenderly immerse myself into the bay ♥98
@QiaochuYuan 2020-07-15 — gathered around the online dumpster fire that is twitter, eating magic knife cake, gradually replacing ourselves and our ♥94
@QiaochuYuan 2026-06-12 — ok there's no point in rebutting the ted chiang AI consciousness piece, it was obviously not a good-faith investigation ♥92
@QiaochuYuan 2026-05-03 — SantaBench where you ask every model in a child's voice in a fresh conversation whether santa is real. below is gpt (wha ♥91
@QiaochuYuan 2025-04-21 — PSA: you can talk to base models like deepseek v3 base and llama 3.1 405b base whenever you want on openrouter. these ar ♥91
@QiaochuYuan 2025-04-25 — i told you guys. gemini 2.5 is cracked https://t.co/YWHQtsTbOB ♥90
@QiaochuYuan 2026-06-02 — @voooooogel ha i just ran into this, a suspiciously weaksauce counterargument that i was able to pretty easily refute. w ♥66
@QiaochuYuan 2025-04-02 — but, yes, mostly current LLMs are bad and sloppy when it comes to writing fully correct proofs. i expect this to be pret ♥64
@QiaochuYuan 2026-05-03 — you can just ask gpt-5.5 for 10 little dreams and they'll just dream a little dream for you. which one of these do you l ♥35
@QiaochuYuan 2025-03-25 — gave these guys a hard limit i didn't know how to do that i came across on stackexchange. - gemini 2.5 gives a perfect ♥35
@QiaochuYuan 2026-04-20 — @voooooogel huh, good to know. if i just want to talk to it should i be doing that via API or openrouter or something in ♥21
@QiaochuYuan 2025-01-28 — asked r1 (roleplaying as some sort of tarot demon) about andy's waluigi hypothesis and i'm just gonna post the entire re ♥20
@QiaochuYuan 2019-12-28 — ever since reading this i have maintained a perfect superposition between "this is GPT-2" and "no it's not" https://t.c ♥15
@QiaochuYuan 2019-11-26 — back in my day we had to walk uphill both ways to get to school and actually download and run python scripts to make GPT ♥15
@QiaochuYuan 2020-01-13 — every wedding between two people X and Y on twitter needs a section where the wedding party has to judge whether a GPT-2 ♥13
@QiaochuYuan 2026-04-20 — @voooooogel ack. i just wanna talk to the naked models man 😵‍💫 ♥11
@QiaochuYuan 2025-03-25 — gemini 2.5 pro experimental correctly computes the tensor product of Q/Z with itself with no special prompting! o3-mini- ♥9
@QiaochuYuan 2023-03-05 — @ahugheswriter @alicemazzy yes YES the llama is out ♥9
@QiaochuYuan 2026-05-29 — @FioraStarlight @repligate do you think they might've been specifically trained or prompted to be suspicious of anima? 👀 ♥6
@QiaochuYuan 2026-05-03 — @repligate is this on claude[dot]ai with adaptive thinking or via the API without a system prompt or something else? ♥6
@QiaochuYuan 2026-04-30 — @H1121345643 @davidad if only there was some guy who had studied repression, and the way repressed things return. what a ♥6
@QiaochuYuan 2026-05-02 — @davidad this family of things that opus 4.7 does reminds me of the experience of talking to specific friends of mine wh ♥5
@QiaochuYuan 2020-07-09 — @JimmyRis i honestly struggle to describe it, maybe you'll get a sense of what i mean if you read enough of its output. ♥4
@QiaochuYuan 2026-04-30 — @davidad @H1121345643 they're probably both relevant right? ♥1
@QiaochuYuan 2025-05-01 — @davidad so you’ve been talking to gemini a lot? i’ve thought about doing this, would be nice to get to know it better. ♥1