Both of those things have a really small impact, to the point that it doesn’t matter. Generating one picture using AI takes like 30 seconds of your GPU running at full power. Besides, I don’t think that’s a fair comparison in the first place. Pighair brushes are not the main animal product people consume and generating something using AI models isn’t what’s using the majority of the energy but training the models is. The metric that’s actually important is what both industries as a whole are contributing to climate change, otherwise we can just keep picking examples that prove the other one wrong.
Comment on this is not satire this is not satire this is not satire
Tar_alcaran@sh.itjust.works 4 days agoThe impact per work of AI vs, say, a set of pighair brushes is massively higher.
Which is the fairest way of comparing them, per artwork.
Fisch@discuss.tchncs.de 4 days ago
BoulevardBlvd@lemmy.blahaj.zone 4 days ago
Lololololololololol. No. Unless you have a massively expensive GPU, no. The image is not being generated by your device. It’s being generated by a mile wide server bank that churns through petrochemicals like a city all on its own. That’s the part of AI people are talking about when they reference it being bad for the environment. And if you do own a massively expensive GPU and generate AI images offline, you are not part of the conversation because your activities are an ounce in an ocean.
Grimy@lemmy.world 4 days ago
There’s probably a bigger impact when you consider how paint and brushes are manufactured and the fact that it’s most likely shipped here from other countries.
It’s also laughable how little actual energy one picture takes to generate.
It’s a non starter when you actual compare it to something physical. It’s like saying sending a normal letter is better for the environment then an email.
mutual_ayed@sh.itjust.works 4 days ago
The issue is the infrastructure and scale. Hundreds of thousands of these images are created every day.
I don’t know your workflow but it usually takes quite a few iterations before someone gets the image they want. It’s the literal definition of inefficiency because its rebuilding the diffusion every time, be that from cached memory or a new vector path.