@maddy you know its a lot of the same for me. I know I had used it to get bits and pieces of things to edit together to make a mockup of my character that I have since comissioned artists to actually make so I can even say that my character is technically the result of early generative AI usage in which I then paid artists to actually draw and create for me. I vehemently reject AI now to the point that I delete programs that just put it in and use worse alternatives just to make a point. But I wasnt always like this and I should not be treated poorly because of something that I had once done years ago.

@maddy oh same here! I think its good to admit that a lot of us had high hopes for this tech. I'm sickened with who is bankrolling it and what they use it for, but I think the tech itself is still really impressive.

A friend of mine once said that the problem is that it creates a sort of mental picket line. Cross the line and you're a scab -- you're taking business away from writers and artists and junior developers. But like... what I actually want isn't for those people to have jobs -- what I want is for them to have their needs met, and a job is just currently a necessary part of that goal. If we had UBI then I think the picket line would be a different story.

the environmental impact thing is still a problem. slop is bad for intersectional reasons.

edit: while I'm talking about LLMs, I want to add that, as far as I understand it, ChatGPT is a fairly bog-standard chatbot wrapped around a kickass search engine, and I'm really pissed about it -- they should just give us that search engine directly!!! I don't want the slop but if they just sold the "thousands of manually categorized pages" part as its own service that would be rad!

@maddy

I used to recommend people Perplexity at my school back when it was new, telling people "it's better than ChatGPT, it has internet access!"

it did not take me too long to resent these companies

@maddy I have used my employer's Github Copilot subscription (we're allowed to use it with our personal Github accounts) and its autocomplete feature for #Kittyscript, my own work in progress programming language.

Pretty much only to save myself the mechanical effort of typing out the characters, which is appealing to me because I never fully learned touch typing because my hands are too big to fit in the base position. I still kept control over the implementation and only accepted suggestions that already were very close to what I had in mind - I did always have a detailed idea of the code I want to write, I was just too lazy to type it from my brain to the keyboard.

I've stopped using it for personal projects entirely now. At work I still use it, although pretty much only the autocompletions. AI use is encouraged by my employer and I don't feel responsible for it.

I don't ask Copilot any questions, I much prefer looking up the actual documentation because that feels (and is) much more reliable.

@maddy i've never used llms (at least not directly, at this point i'm sure something i've used has a slop machine embedded), but i was interested in running a local model and seeing what it could do when stable diffusion became available

i didn't only because i was lazy and i wasn't that interested in it. never been interested in chatgpt or any of the cloud services because i have a long-standing skepticism of cloud services in general

and i also have longstanding hatred of talking to computers using natural language (predating llms). so llms would never be a technology i'd get into

as i learned more and more about it i quickly started to despise the technology and now i'm doing all i can to never touch it

that said, i don't have any smoke for people who tried it out in the early days when there was a lot of hype and most of the downsides weren't obvious at that point

even for people who used llms more recently, i won't judge as long as they've stopped using them. and i won't judge people who are forced to use them at work but don't use it beyond that