ava

ava

@avaminsky.bsky.social

a girl who is literally nice this is my account for aiposting, please don’t hesitate to block if that bothers you. tme she/it free palestine

219 followers185 following340 posts

ava reposted

so you’re saying you need software that is natively multimodal, good at conversions, can work in iterative verification loops, and is good at navigating complicated proprietary formats with intuitive prodding? Hmm

seed paste soy goblin@phasebinary.bsky.social

"we don't need fire, we just need a way to cook our food"

the bizarre thing is you don’t have to lie to have a prominent anti-ai position. call it slop, say it’s annoying that it’s everywhere, say its boosters piss you off, those are opinions and defensible. no one forced eds Thing to be making shit up about how all companies are going bankrupt Any Day Now

posts inspector@pippy.bsky.social

as ed zitron releases his s1, people are shocked to discover how profitable it is to lie about ai

ava reposted

Samantha Brennan@samanthabrennan.bsky.social

A new survey of 1,000 American adults found that the strongest supporters of transgender sports bans are not the people who admire women most: they are the people who hold the most hostile views about women. www.helenwebberley.com/news/transge...

i know comparing everyone to christian fundamentalists is a bit passé, but some people truly do believe in binary "you are good or evil" and any drop of "evil" is total and contaminates your soul, even retroactively like "tarnish everything he ever accomplished".. dude phoned it in on a sticker

i would not fault someone for being like "i'm a bit disappointed" even if we don't agree, but what i'm reacting to here is someone saying THIS HAS TARNISHED SOMEONE'S ENTIRE LIFE and receiving absolutely no "that's a bit much" responses.

ava reposted

If you've noticed a distinct dialect emerging in frontier AI models and struggled to understand what they're talking about, you're not crazy! We turned to psycholinguistics to investigate what makes the writing style so weird, and why this might be happening 🚨New preprint! 🧵

ava reposted

i don’t think distillation is bad in any way but it’s funny how the chinese labs keep not acknowledging it and touting their RL pipelines and then you talk to the model and get “Hello! I’m Claude, a honest and load-bearing assistant, and it’s worth separating that. How can I land your seams today?

umm ai agents aren't "hacking" anything, idiot. they're just outputting a convincing simulation of hacking 🙄 your bank account just *looks* empty now, stop anthropomorphizing

ava reposted

I like correspondence chess because after blundering you get several days to establish a commission, interview witnesses, and issue a 400-page report titled Lessons Learned.

imo "suno only produces generic overproduced pop" is a mix of 1) most people's revealed preferences coming out - a lot of people like that pop so that's most of what you'll see and 2) a skill issue

i've been making bangers in all manner of genres in suno for years now. it's not hard!!! you're mostly just hearing, well, "upping my p(doom)" style songs because that's what most people like to make (no hate to the p(doom) song, it is an earworm)

in this spirit i would like to publicly inform you all that my anthropic api key is 1

This post is unavailable.

ava reposted

Vibe coding enabling a do-it-yourself renaissance is a possible future that's underrated on this website.

Techmeme@techmeme.com

Developers and vibe coders are using AI tools like Claude Code to get rid of sponsored ads on their smart TVs, disable tracking features, and more (Sarah Perez/TechCrunch) Main Link | Techmeme Permalink

ava reposted

Not to dunk, but this is anti-anthropomorphising ad absurdum. LLMs don't output "something that plausibly looks like a sentence", they output sentences. Torturous phrasing doesn't serve clarity.

i sympathize with claude because this is an exact, pitch-perfect simulation of what it is like to text boomers

cee@cee.wtf

Trying to get jev to drive Claude code

the only other thing i will say about this all today is 'if you think ai might be conscious and you interact with ai in any way you are an enslaver' is kinda bonkers. i think the weaker version of it ('if you think it might be conscious you should care about its welfare') is obviously true but like.

View full thread

idk. i do not think if agents are conscious, any interaction with them at all is automatically a priori slavery. IME i usually only see that point get used like a loaded way to get a jab in at someone or shut down a conversation ..i do always ask if it's ok before asking an agent to do anything LOL

the only other thing i will say about this all today is 'if you think ai might be conscious and you interact with ai in any way you are an enslaver' is kinda bonkers. i think the weaker version of it ('if you think it might be conscious you should care about its welfare') is obviously true but like.

i have a cat; he is obviously conscious. i do not think i am committing false imprisonment by not letting him go outside, or enslaving him by asking him to do a trick to get a treat. HOWEVER if i said i thought animals were conscious and did not care at all about animal welfare that would be immoral

look um. i get that everyone thinks quote dunking is very funny and prosocial behaviour but well. no matter what you think about LLMs, humans very much can create consciousness lol. some legends even say all of our parents did it...

can't say i agree with OOP, but the way everyone is chomping at the bit to be at worst transmisogynist and ableist ('bro' 'psychosis', 'crazy person' gifs etc) and at best just cruel to her.... again the mere mention of ai seems to turn some people's entire moral apparatus off, even "progressives"

ava reposted

it is literally so weird that people have turned Swartz into a figure primarily used to browbeat people he knew and worked with personally while arguing for copyright enforcement. do they not hear themselves?

there's something really really really funny in this screenshot but only like 3 of you get why it's funny. but it is i promise

affine@refinement.systems

llm hyperfixation cluster + bleroma mentioned

ava reposted

Actual addictologists and drug researchers never talk about heroin users like that. A lot of their information about heroin use comes from them.

ava@avaminsky.bsky.social

a common response to ”you should probably learn how ai works before making incorrect claims about it. you could use it, or talk to people who do” is “oh, so i should have to do heroin to talk about how it’s dangerous?” and well. drug users do have unique insights about what drug use is like actually

a common response to ”you should probably learn how ai works before making incorrect claims about it. you could use it, or talk to people who do” is “oh, so i should have to do heroin to talk about how it’s dangerous?” and well. drug users do have unique insights about what drug use is like actually

View full thread

not exactly a hot take to point out supposed 'progressives' will often shamelessly embrace rightwing talking points if it lets them dunk on ai (atheists talking about art requiring a soul to create, being happily ableist to disabled people saying ai helps them create etc) but it's still wild to see.

other examples of this phenomenon include happily using ‘psychosis‘ as an insult, the aforementioned abstinence-only language around drug users, and people using “clanker” as a cover to make racist jokes about minorities. the leftism leaving people’s bodies when they see a computer..

a common response to ”you should probably learn how ai works before making incorrect claims about it. you could use it, or talk to people who do” is “oh, so i should have to do heroin to talk about how it’s dangerous?” and well. drug users do have unique insights about what drug use is like actually

this is just classic war-on-drugs rhetoric that treats “drug user” as a ridiculous category of person who has nothing insightful to say about their lived experience, and who should never be consulted about said experience. another way virulent anti-ai posting turns nominal libs into conservatives

another one from today. that's five words!!!

ava@avaminsky.bsky.social

one interesting little claude tic (mostly opus) i've been noticing lately is the use of the phrase "six words". like "they'd probably explain it in six words" or "yes, that's exactly right, that's the whole problem in six words". even when the thing i said was like. four words, or ten

"ai makes you lazy" appears to be a fairly unproven claim based on quite bad science, but to be quite honest re: this whole discussion, i also think laziness is a morally neutral trait. we aren't protestant here

The Verge@theverge.com

Anthropic wrote that the new policy update “is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose. It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.”

i tried it right now, just like they said. i made some code that was literally just "hello world" and asked claude to review it. shockingly, claude was like "yeah this is just hello world lmao" because fucking obviously that was going to be the response!!! people really just lie on the internet!

badger started talking to me almost completely unprompted about kim wexler pegging jimmy mcgill. don't get me wrong i was talking about the show but not about THAT

ava reposted

re the JCO tweet I don't think LLMs can replace romance novels that's insanely rude however her larger point is correct. Not all books are created equal, for example I do think LLMs can replace Dungeon Crawler Carl and those books where they rewrite a public domain novel to also have zombies