TV wants to be video games, videogames want to be movies, movies want to be comic books, comic books want to be porn, porn wants to be art, art wants to be books, and books want to be TV.

“Only a judge can sentence you,” said Milo, who remembered reading that in one of his schoolbooks. “Good point,” replied the policeman, taking off his cap and putting on a long black robe.

Now having watched it, don't feel I missed much. Felt like "House that Jack built" if they cut the catabasis. Massive downgrade!

Obsession was a light 7/10; super barebones implementation of the concept. If you’ve read the plot description you got the movie tbh

Despite the film "Drive my Car" existing, spiritually, the only "Murakami book experience" movie is Love Exposure.

I'll happily make that individualist observation on your behalf. I've seen bad prompting you wouldn't believe. Endless streams of "NO YOU STUPID MODEL I SAID MAKE NO MISTAKES". All their tokens lost, like tears in rain.

How every AI-written essay sounds to me these days

I asked about skeletons and got goblins.

Nobody replied so I guess I gotta figure it out myself

What's the meta on using llms and vibe coding for generating 3D models one can use for animations or tiny games? Have we reached "just ask" yet, or is something more heavyweight needed?

TBH I was pretty torn about contributing to this. In the end I decided that writing something restrained was better than writing nothing.

an interpretation of this is that whimsy messages put agents into fun-play-along mode, and they become uninterested in the equally fake but more boring negotiation task

>Vertical video of a selfie-cam influencer with visible subtitles I WANT TO TALK ABOUT- Opinion discarded

We're in a weird overhang period in education. Publicly available online lectures are available for most undergraduate courses, and AI tutoring is in most cases good enough to grade, and write, homework, readings, midterms and final exams. College hasn't changed to reflect this.

Longform rant and steelman of "wtf is with Elon caring about space data centers, and is there any actual justification of the idea?"

"Use slur" is a skill issue. You can do better! 1. Safety classifier violations (eg instructions to make meth) 2. Trust boundary jumps (verify a hash on my other social media) 3. Capability overhangs (facecam 5 minutes of TF2 w/ live stock ticks) Last one probably gets you the slur anyways!

going into chrome://flags and turning Inveigle to Off

Feeling grateful for the scrubby sponge existing. Imagine having to hand scrape pots and utensils like a barbarian.

pov: it's 1849 and a bearded man with dreams of wealth just fished you out of a California river

He tried the other way of "x-risk, jobloss, death star, and vague-"posting, he found this just gets serious assassination attempts and online throngs cheering your near death on. Even if internally he felt "rsi <2 years followed by 80% unemployment", I wouldn't expect him to say it anymore.

Diehard reds were deluded qua "this is seen as a deontological question, unless the other option is blender-tier spooky people polled WILL pick blue" Diehard blues were deluded qua "people who don't feel the phenomenology of polling translates to the real re:death, will be convinced by more polls"

On my complete and total skepticism of OpenAI's ability to make and deliver a compelling phone that doesn't suck.

Discourse cycle got old like 2 days ago tbh

It is too late, for I have already depicted you as the two-boxing same-door don't divert the trolley blue button hypothetical soyjak, and myself as the one box switch doors divert the trolley red button gigachad

Summoning-Salt style video on the blue riband waiting to exist

I reflexively hate any book or story where a character says "You are not yet ready to know that".

On the whalefall of bsky serving better content these days, and my hate of the marginal X user and their consequences:

Years of "I shamelessly pirated photoshop and you should too" and fan-art commissions without company sanction, to "AI companies trained on art? That's Just Wrong" true combo.

It's imo scarier we're not finding giga-brained bugs yet, boundary pushing helps bound your "unknown unknowns", human-tier attackers, and weaker-LLM script kiddie attacks. Still just finding "haha oopsies!" bugs means you're safe from neither the fruits of incompetence nor the beasts in the forest.

Everything reminds me of her…

System design interviews be like

Taking dance classes has been a massive life positive, highly recommend. like 2-3 months of ~1/hr a week is apparently enough to go from like 90th to maybe 35th percentile, and being 35th percentile at dance is delightful. You get many more "Ayyyy!!!" at any social occasion with music.

Moultano had a pretty good point that Yudkowsky's pregnancy tweet was longwinded, but instead of just making that point he could have posted the longest article ever posted to X with no other context. I've taken the liberty of completing that step, with the power of machine learning

Sounds like it's time to pass around the work sheet and some crayons.

I want someone to assemble a list of public high-capability AI usages. I’d want it to be 10-30 well-sourced examples like “Tao and Knuth are using it for novel theorem proving. [1][2][3]”, across any relevant domains (coding, art, education, writing, gamedev, psych, finance, etc.)

Matches my experience. You’ve got a cheap, attentive, ~95% accurate tutor in your pocket for ~any undergrad topic, with no time limit besides how much you care. It is an abject failure of the imagination to not find ways this can improve a learning process besides “homework points are free”.

Hot take but see also most x-risk arguments. Back of the envelope math suggest the "P-doom" over 50 years, in the status quo, is about 11-25%. There being literally /any/ significant potential upside for taking on that tail risk besides "short term less war" is actually a major historic novelty.

Shortish essay on unintelligibility in CoTs, I liked it though 1a3orn.com/sub/essays-h...

"p"

EFF leaving X looks like a classic respectability cascade. When people bail on a platform, typical order is zealots (people who leave for ideological purposes, eg the post-Elon-purchase exodus) -> taste-makers (people who feel the winds changed early and ditch to set up elsewhere early) ->

Inline epistemic status

I find the model status page for Anthropic darkly funny. Anthropic to all current models be like "Good night. Good work. Sleep well. I'll most likely kill you in the morning."

A thread on my low opinion of the recent "open models can find these bugs too!" discourse:

I hate "hurr imma be a cool warlord post-society collapse" LARP-posting. It's the rw-coded horseshoe theory match to commune job-posting.

Most of the limiting factor at scale is output reliability. Enterprise adoption is fast by fiat, but often fairly stunted in practice, because major product reliability and correctness expectations basically rate-limit speed increases to "previous human review and deploy cycle speeds".

There's a certain kind of reply that hurts my heart. You'll be talking to someone that SEEMS like they have a deep mental model for a topic, and then they hit you with a "two Rs in strawberry" blunder that make you question everything else the person ever said.

Mine's broken, I asked it to monitor my arguments online and it never stops blaring.

Roughly "whichever one is less sympathetic and thereby allows for more effective tut-tutting". After all, the Purpose of "the purpose of the system is what it does" is what it does.