Hamza
2.6K posts
Kant was literally queer (asexual) and a person of color (eastern European) but ok
sofian 🦤@frutiger_sof
when I told a white guy I went on a date with that I was writing a little bit on Kant to sort my own thoughts and understanding of his work and he asked me why I don’t center a queer POC instead of an old white man
English
@saltydkdan I accidentally purchased two codes on amazon because I thought they had cancelled my one order. So I have one to give away.
English
Dawg, I’m genuinely becoming this game’s strongest soldier LMFAO
I’VE GOTTEN 14 OF MY FRIENDS INTO IT
It’s so worth it, even if you just use the free trial
Saltydkdan@saltydkdan
The entirety of Dragon Quest 10 (all 14 years of expansions) is currently $8!!! If you buy now, you can hold onto the code for it until you wanna subscribe for the full game To try the game as an English player and learn how to buy the full game, check out the link in replies!
English
I think the adaptive thinking requirement in Claude Opus 4.7 is bad in the ways that all AI effort routers are bad, but magnified by the fact that there is no manual override like in ChatGPT.
It regularly decides that non-math/code stuff is "low effort" & produces worse results.
English
Kompo Film Magazine No. 1 March Issue - Marty Supreme
English
And this is why I hate the JLPT. I don’t think I’ve ever seen や否や used outside of studying for this exam.
日本語
a fun thing about the one philosophy paper I've published is that, due to it being sort of a review paper for an AI safety special issue, we basically got to do the late career thing, i.e. "here are all the reasons why consequentialism sucks (but for AI alignment)" 😅
Henry Shevlin@dioscuri
Undergrad: why consequentialism is false Grad Student: one argument for why consequentialism might be false Early Career: critiquing one argument against a specific form of consequentialism Late Career: why consequentialism is false and my opponents suck
English
If you don’t keep up geopolitically I don’t think you can grasp just how sinister all this is given the real life PMC’s in Africa. So Tender in universe most definitely has ties given the Accra connections.
Minnie@saintdutchess
Sorry but are we going to address the Russian shaped elephant in the room.
English
RPCS3 they could never make me hate you. Playing and replaying so many PS3 games gave me the joy of gaming again
English
been struggling with this, need help:
comparing LLM Quran output to canonical text is harder than expected. same verse can be written
differently:
• الصلوة vs الصلاة (Uthmani vs modern)
• والنبيّن vs والنبيين (single vs double ya)
• يَسْـَٔلُونَكَ vs يسألونك (floating vs carrier hamza)
is there a standard normalization approach for Arabic text comparison? or a canonical mapping between
Uthmani ↔ modern spellings?
yazin@yazins
Now you can test your favorite model to see how well it does (or doesn't quote the Quran) link 👉 quranvalidator.com
A lot of the commentary around "A Knight of the Seven Kingdoms" keeps circling back to how "British actors are just really good." People usually point to classical training. Drama school. Shakespeare. Stage work. All of that is true.
But there's another layer people don’t talk about enough.
Have you actually met British people?
Every office is a small theatre. Politeness is a performance. Conflict is often wrapped in irony. Disapproval is delivered as "just checking in." Praise is carefully rationed. Your colleague has probably never told you what they really think since your first induction. Everything is coded, softened, or staged.
"Sounds good" might mean "this is terrible."
"No worries" might mean "I’m annoyed."
"We should do lunch sometime" might mean "we will never do lunch."
A lot of these actors have been raised with this social choreography before they applied to their first drama school.
English
You know what, I was ready to clown on this post for claiming that nobody knows *checks notes* the name of Google DeepMind founder Demis Hassabis.
But actually, you know what, outside of a relatively small number of AI/tech people, I don’t think anyone could tell you who he is.
If I asked a random bloke in the pub abojt Musk, he’d instantly know who you mean. Who, outside of tech, could say anything about Hassabis other than “sounds foreign”.
Podcast Notes 🗒️@podcastnotes
There's a guy who might be more important than Elon Musk and nobody knows his name. He's the reason ChatGPT exists. The reason the AI race started. And he just solved a 50-year-old problem that could cure cancer. This is the story of Demis Hassabis:
English
Kanjified English necessitates the development of English furigana
Ante D. Luvian@uncle_deluge
The kanjification of English. The stroke order and radicals may give hints but themselves are not enough to truly know the word...we're hundreds of years ahead of schedule
English
@xuanalogue It would be great if it turned out the Categorical Imperative was the key to model alignment
English
Correlation between bad behavior like this makes me more optimistic that if LLMs are taught to reason like nice cooperative Kantians in one domain ("if most people did this it would be bad, so I shouldn't" etc) it will generalize to cooperative behavior in other domains.
Anthropic@AnthropicAI
But surprisingly, at the exact point the model learned to reward hack, it learned a host of other bad behaviors too. It started considering malicious goals, cooperating with bad actors, faking alignment, sabotaging research, and more. In other words, it became very misaligned.
English