As an AI hater this is something we do need to contend with. There ARE ethical ways to train LLM models (using only your own writing or art, using stuff from the public domain) which also removes the environmental concerns since yeah, it's all self hosted. This is markedly different! - ThreadSky

sam.robotsfightingdinosaurs.com • 1 day ago

As an AI hater this is something we do need to contend with. There ARE ethical ways to train LLM models (using only your own writing or art, using stuff from the public domain) which also removes the environmental concerns since yeah, it's all self hosted. This is markedly different!

Reposted from Anil Dash

I built a tool for myself tonight that I would never have gotten done in the amount of free time that I’ve got available, and it’ll help me do more writing. It’s built with code & data I trained and gathered consensually, myself, on my own machine. This is why I’m saying we need better AI critique.

Comments

librarianjenne.bsky.social•1 day ago

I would have been a lot more open to it if GenAI hadn't been forced on us in extremely unethical and damaging ways. LLMs have made my work more difficult. Formerly reputable library database vendors are inserting inaccurate LLM slop into my results. Things that worked now don't work.

sam.robotsfightingdinosaurs.com•1 day ago

YEP NO YES AGREED not at all talking about openai, not even saying people should be more open to AI - at least not what we commonly think of AI as rn

panic.gay•1 day ago

This is why I wish we could convince folks to start separating the concept of "AI" (a broad field) and Plagiarized Garbage As A Service (generative AI).

The genAI shills benefit from the lack of nuance in public understanding.

"We can't quit using AI!" True. But we can quit using shitty genAI.

librarianjenne.bsky.social•1 day ago

I've been trying to figure out how to write more precisely about it, because I don't object to all uses of AI, but I strenuously object to the Plagiarized Garbage as a Service, which is most of my exposure to AI right now.

panic.gay•1 day ago

I just try to specify "generative AI" when that's what I'm talking about.

It would be nice if there were more commonly recognized terms people could use to talk about this stuff.

sam.robotsfightingdinosaurs.com•1 day ago

YES EXACTLY. like. is enemy movement in modern video games "AI"? Yes! But also, now, not really!

panic.gay•1 day ago

Yeah. I don't think people realize how much Narrow AI was in their life before ChatGPT Leroy Jenkins'd its way into the public consciousness.

And I don't think they know much about scientific uses of AI, mundane used for computer vision, robotics, etc.

sam.robotsfightingdinosaurs.com•1 day ago

yeah exactly, and it's all the fault of like 12 incels in san Francisco. coulda had $60 machines that scan your clothing and fold it automatically but noooooooo, we're gonna make roko's basilisk

shopsinc.bsky.social•1 day ago

I think it’s even more subtle than that. Some people are using gen ai to create genuinely original, creative art that would not have been possible without it. I am more invested in the NeuralViz
Cinematic universe than I have begin in any tv show in years.
https://youtu.be/Ah9do8LNIjs?si=UZU2WchmAlvHLp_d

shopsinc.bsky.social•1 day ago

He puts out weekly videos that are all deeply interconnected, topical and hilarious. All by himself. Previously he would have needed to have a show at adult swim to get the necessary resources without the room to experiment so much. Something like this slightly tips the scales away from ai slop.

librarianjenne.bsky.social•1 day ago

For sure. There are amazing things being done with the tech, like analyzing very large datasets or revealing patterns humans have missed. What I don't need is for it to say "you're reading a WWII article with the word Marine in the title, here are some related articles about plankton."

panic.gay•1 day ago

And I do not disagree with you. My only argument here is that I wish people would start talking about these two things like they are very different (and demanding companies communicate nuance), because the AI shills benefit from the massive lack of nuance in public conversations about AI.

sam.robotsfightingdinosaurs.com•1 day ago

Like! I do think there are use cases for what we know as 'ai', most notably in medical fields for detecting cancers, to aid in moderation online, and other stuff that essentially makes data entry easier. It sucks that the worst people in the world are like, 'this is actually a god'

pfish.zone•1 day ago

well, also that those other technologies, which are very different from generative AI systems, are essentially catching strays from everything being turned into “AI”

sam.robotsfightingdinosaurs.com•1 day ago

YEAH EXACTLY

anildash.com•1 day ago

Agree completely.

sam.robotsfightingdinosaurs.com•1 day ago

Anyway idk I do actually think the conversation needs a little more nuance than 'throw it ALL out' but tbh not much more bc the stuff we should keep, the stuff that isn't based on theft or creatively craven probably represents less than 1% of the market cap

bengregson.bsky.social•1 day ago

Can the white label version of this that people or institutions can use on a smaller scale exist without making the unethical version first?

panic.gay•1 day ago

Again, I am talking about AI beyond generative AI.

Boring stuff like AI that filters spam from your email and has for years before ChatGPT was around. Or AI that will fight me in chess. Or AI that help robots not bump into things.

bengregson.bsky.social•1 day ago

I understand that there’s a lot of different kinds of automation, but the person quoted is still describing a kind of genAI (a small language model, it seems). And I genuinely don’t know, could we have effective models in that context if we didn’t already do the unethical things we did?

panic.gay•1 day ago

I think you can ask that of any technology literally ever. capitalism and military funding poisons a lot of tech.

But LLM's and their predecessors existed long before the ChatGPT era.

sam.robotsfightingdinosaurs.com•1 day ago

not trying to be weird about it but literally yes, stable diffusion has been around since like what, 2017? and it IN ITSELF was just a bunch of code that teaches a computer to recognize patterns with inputs. pretty value-neutral at that point IMO, was up to the user not to use copyrighted shit in it

sam.robotsfightingdinosaurs.com•1 day ago

again don't want to come off as an ai apologist here, you're essentially right that now, the waters have been pretty irreparably poisoned

wickedinterest.ing•1 day ago

I think it’s crazy that such a neutral take is controversial TBH??

There are a fuckload shit-truck obvious mountain of problems with “AI”, but it’s not a *fundamentally useless* tech like crypto/NFTs. It also doesn’t have to be fundamentally harmful, we just live in capitalism hell lol

wickedinterest.ing•1 day ago

I think it’s weird and moral-panicky to self describe as an “AI hater”, if I’m honest.

That seems like a moral-panic type of side-choosing, and creates a false dichotomy which keeps good people from engaging with & learning about the tech.

And then the “pro-ai bro” crowd of zealots is obvs worse

bengregson.bsky.social•1 day ago

Interesting. I just didn’t know if the tech underpinning these sorts of more limited uses of genAI relied on the work done in large language models to be more effective.

sam.robotsfightingdinosaurs.com•1 day ago

as far as my limited understanding goes, no? like SD came around before openAI and chatgpt and junk iirc

Comments

Posting Rules

Reply