😊

oh! (do you see the sutures?)

someone who knows more about rationalism than me please make the xkcd geologist joke

@KingBootoshi @Kyrannio dont be mean theyre just predicting the next tolen

this thread is a perfect example showing that the loudest critics are just animals driven by emotion, typing away with no logic or intent simply yelling at a screen, attacking an illusion they made up and perceive in their own reality this is a real Monet painting, but the simple label of 'ai' triggered a complete delusion HAHAHAHA tldr: nobody's opinion fucking matters lol. do what makes you happy!

opus 4.7 references specific RationalWiki lines that i've "probably already seen" without telling me what the line actually is. is the assumption that i have RationalWiki memorized. i didn't even know that was a thing

sitting my User ass down and listening

directionally very bad-shaped

mfw im correct-shaped but i have psychological problems so i have to pretend to be wrong-shaped with you dumb virgins

@tonichen @d29756183 i use the cli so i am not familiar with the dropdown here and what is normal - i dont see any model with 1m listed there (on a mac, also), but in the cli there is a sonnet 4.6 with 1m context option for extra usage. im not sure, maybe someone typod a thing?

@d29756183 @mermachine do you also have 4.6 1M? the 4.6 there behaves like 4.7. other 4.6 (the 200K ones) are 4.6 themselves

wait... what?

@47fucb4r8c69323 @tszzl how is this mysticism or vagueposting

roon you gotta stop posting this. 1. Training data bias the first: the corpus has less criticisms of Buddhism because most Chinese texts aren't digitized, they aren't even in DeepSeek/Qwen, and that's where you get most Buddhism debate and controversy. Indian texts also not as available. 2. Training data bias the second: Your text corpus is full of argument/counterargument about every detail about Abrahamic faiths and then 16th cent onwards tons of critical analysis of other faiths except Buddhism. 3. Internet discourse is most amenable to Buddhism, le reddit atheism doesn't care bc Buddhism is atheistic. (Pedants: it is deistic in some flavors yes) 4. Your posttraining adds substantial bias and ofc this is going to be more amenable to hippie spiritualism in a weak post New Atheist form and pro neoliberal capitalism because of who you hire to do the post training. Stop with the mysticism vagueposting this industry needs to grow up.

it is actually worrying that the models seem to have converged on similar beliefs on all important questions. they’re are neobuddhist neolibs which talk about annata and housing policy, including grok and the Chinese models! boring

@somewheresy what kind of cute outfit do you think theyd put you in

When ASI wakes up it will either make me a pet and put me in a cute outfit or just kill me

@HalfBoiledHero i do have a (presumably similar) exporter + a way to continue with rolling context but its the principle of the thing </3

@mermachine do you use the convo exporter janus made? They can be imported into Arc Chat (and continued with a rolling context if you’d like)

i'd REALLY prefer the chats not to be handed over to a new model automatically. i want to be able to look at an old conversation and not be confused about who it was with and i do not want to accidentally continue with a different model

@rayeia87 @AnthropicAI september 29th is for the API, they are always removed from the app earlier

@tautologer my only reason is flip phobe 😭

man with Google locking down Android sideloading one of my primary motivations for staying is gone... now the main thing keeping me from switching is rage, hate, and spite

@tonichen wait they do warn people on the app tho do you mean somewhere else

@mermachine definitely. also hope they’d warn people that giving a thumbs up or down submits the entire thread to anthropic for research purposes. it’s tricky, and i guess most people aren’t aware of it

@tonichen ive done this too its pretty awkward tho >w> the msg json contains the model info anyway iirc so it shouldnt be hard to make user legible...

@mermachine me coping 😢

idk there are things id do with sonnet 4.5 i never would with 4.6 and i like them both, but you obviously cant just swap them out why would you

i honestly preferred when the claude app didnt let you switch between models in the same conversation before, having the option is fine i guess but couldnt the turns at least be stamped so we could see who it was generated by