It's legitimately no different than cryptocurrency and NFTs, but this is the techbro scam that finally stuck after flinging shit against the wall over and over.

The training data sources are built destructively, without any consent from the artists, writers, programmers, anyone whose data is pulled into these models. The information these things spit out usually has so many inaccuracies - search engines are fucked, software is built without understanding a damn thing it actually does, poor translations of documentation are made public, the list goes on.

On top of that, data centers are one of the biggest environmental nightmares we've ever seen. Why are we pushing these so hard in the middle of a goddamn climate crisis?

There's genuinely not a single thing it actually does well, and the only reason it genuinely took off is because investors in the same 5 companies are patting each other on the back.

It's also worth mentioning how chatbots built off this stuff are designed to always be extraordinarily friendly and confirming to the user. It's such a massive factor contributing to AI psychosis, to the point that there people who legitimately believe these things are sentient and conscience. Meanwhile we've already seen just how easily these closed source models and be tweaked to spit out pure propaganda, making them an extremely valuable political tool - it's no wonder fascists love this shit so much.

I'm a graphic designer.

The entire point of AI as it exists in the western imperialist world is to cut out workers, labor rights, and the means of production by automating jobs, capturing processing power to replace brain power, and essentially to funnel money from consumers to those who control AI. Those companies are using our materials, our work, our capital, and our attention to train their models, and NONE of it is meant to make our material conditions better. What do we get from it now? Text that sounds like a middle manager, and that's factually correct as often as one. Images that look like shit, and that are making people dumber. Cars that disable their AI at the last possible moment specifically to avoid responsibility, and that doesn't work well-enough to be road legal under ideal circumstances. Search engine answers that show nothing but that AI has automated being confidently wrong on the internet. Entire towns that no longer have water or electricity, but do have constant jet noise, diesel pollution and markedly higher temperatures.

We aren't even at a stage where anyone but Nvidia is actually getting anything out of it. Budgets aren't being saved because companies are using AI, they're just spending the money on AI instead of for actual work. It's a cult. It's operating as a cult, and for some reason, no one is treating it as if it's a cult except "Rationalists", who are balls deep in it because it's a cult.

lmao, honestly from what i've seen lemmy is very much pro slop -- and it's been quite disheartening. i've mostly just stayed on mastodon lately cause i don't have the bandwidth to fight with people who say "but local models!" ... nah fuck that.
LLM's are a racist, fascist, anti-human, anti-social, anti-intellectual, regressive, inefficient, ecoterrorist, counter-revolutionary, class traitor technology. Even with open weights / open models ran locally, it still only reduces peoples capacity to learn, create and build community. It still pushes us as a society apart. There are technologies that can be up lifting and constructive, but this isn't one. There is no reason to chase more lines of code, or more pages in a book -- instead chasing to better understand, better communicate, better organize together.
Every "leftist" person on lemmy who is pro local llm's, is just excited about the technology and that leaves very little room for organizing and utilizing this unique time where there could be a non-partisan reason to organize against the oligarchs and fascists who are in power in much of the world right now.

I dabbled with AI usage for a while, ran some local models, would try whatever was the current best for when there was a project that I didn't really want to put the effort into learning about--and I found it was just always regressive and annoying. I'd brick hardware, or my context would get so big that the output would continue to become more sloppy.
I started just going back to reading documentation, and it was just always easier. It was always easier to just grep for whatever I needed. Whenever I'd go back to simply trying to understand the problem and just fucking read, I'd feel almost like a haze being lifted.
I think this is the point where I realized, most complaints that people have with software etc are already solved issues that most "tech" refuses to implement or has enshitified. So instead of just solving the problem in a predictive / deterministic way, we are instead being sold a technology that will burn incredible amounts of fuel, will requires incredible amounts of resources all to what--search through my email? More slowly than what was developed in the 70's would.
Then I read 'Blood in the machine' and 'End times fascism', watched 'ghost in the machine' and read more most and articles... The sheer mass amount of violence and suffering created is unimaginable.

I think my stance is mostly, even without the ethical issues -- what use would this technology even have? Why would I want a computer to do the learning and creating, when I could instead enjoy the work of my friends, and enjoy creating something for others; while building a community. and community is the #1 way to curb fascism.
also my computing is more and more evolving into something more similar to permacomputing. My music and art taste has always been more into finding & connecting with artists locally, or at least artists who I can actually interact with and understand. Outsourcing these things to a machine would not bring any joy.