Every now and then I see someone complain about fedi being "too AI hostile" but everywhere else I go is *completely deluged* in slop. It's not like it doesn't happen here too but it's nice to be somewhere that feels like a break from all that
1 Boost
1 Like
@cwebber I actually like quite a lot of AI generated images. Just not the ones that were made for a specific purpose since they are soo bloody boring and bland. Generative systems can yield awesome results, especially if you like strange and weird images that look like the hallucinations of an alien mind or surreal body horror and use models that are still rough around the edges. It helps if you write prompts that don't say anything about the contents of the image but a lot about the style and/or mood. Using random strings for prompts works nicely with those simpler models, too.
The only thing I hate about AI is the commercial AI companies and the hyperscalers buying all the computer hardware for their "clouds". If it weren't for them, I could afford the hardware to run all the AI software I want to play with locally, no computing centre needed, not even a network connection. All I want is some random images conjured from latent space, the weirder the better. And perhaps also some AI music, although not the polished high-end stuff, but strange audio extrusions that vaguely resemble actual human music. Like the stuff the old Riffusion produced, like the endless AI music streams from Dadabots many years ago, which sounded like strange alien approximationn of human music. I loved it when GenAI was still deep inside Uncanny Valley because I love it there, I want to move to Uncanny Valley and marry a robot. Nowadays GenAI has moved out of Uncanny Valley into Boringville, it's almost indistinguishable from human-made stuff, yet far too smooth and homogeneous to be interesting.
@cwebber
Is there such a thing as 'too' AI hostile, though?
@cwebber How about Wikipedia? En.wikipedia *is* deluged in AI slop but has become increasingly strict against LLM usage, apart from very limited exceptions. There's a noticeboard https://en.wikipedia.org/wiki/WP:AINB and LLM warnings are pasted in articles that are suspected of being LLM-contaminated.
@boud Wikipedia's anti-genAI policies are good and people are doing a good work de-sloppifying wherever they can
@cwebber And most "anti-AI" people aren't ranting, they are just being honest and realistic. You don't have to care about the ethical and ecological issues[*], you don't have to have magic future vision..to see that 1) slop sucks and 2) AI is a direct, personal threat to your livelihood.
It's the "tribal" attitudes currently in vogue: if you're not all-in, you're the enemy and my team must win."
(* - although you should)
@cwebber Poor sloppers only have 99/100 places on the Internet. So, soo oppressed. :(
@cwebber My BSky timeline is *full* of slop. It's exhausting.
@socketwench Are you telling me that fedi doesn't need its own Attie to succeed
@cwebber @SteveTux At one point I definitely saw a lot of completely nonsensical hate for anything that had the word AI on it, no matter what it was, and I was getting a bit worried because while that was happening in the background, I was using LLMs to describe photos that were missing alt text and help me fix code that had major accessibility issues in it. But everyone has calmed down a lot. I see more criticism of the way AI is used and marketed than of the entire concept now.