Image generation is often done remotely, using massive amounts of energy and water cooling. I enjoy the funny images as well, but I don’t like the massive volume of AI images that make it tougher to find human artists.
That’s just nonsense they made up, or they are confusing their anti-AI memes.
An image generator uses no more power than a video game. Most run on consumer graphics cards.
LLMs are the ones that use a lot of resources but “massive amounts” is a vague term that has no useful meaning. No matter how you try to refute their statement, they can always just declare that they mean something else.
Generating the image does not take much power, though training them does; I should have been clearer.
When I say “massive amounts,” I mean a company like Microsoft opening large data centers that require enough energy and water to disrupt local communities. Obviously this isn’t an AI issue, and Microsoft doesn’t train for image generation AFAIK, but the fact remains that training an AI model requires an order of magnitude of more resources than most consumer or corporate applications.
If AI models were only getting more efficient, I wouldn’t worry about this, but companies tend to scale up and use more resources to make larger models.
![[Rusty Creates] 'Artists'](https://lemmy.world/pictrs/image/ee6ccbcc-d001-4b28-a88f-d39c2b75f860.png){kind=link}
Image generation is often done remotely, using massive amounts of energy and water cooling. I enjoy the funny images as well, but I don’t like the massive volume of AI images that make it tougher to find human artists.
Removed by mod
That’s just nonsense they made up, or they are confusing their anti-AI memes.
An image generator uses no more power than a video game. Most run on consumer graphics cards.
LLMs are the ones that use a lot of resources but “massive amounts” is a vague term that has no useful meaning. No matter how you try to refute their statement, they can always just declare that they mean something else.
Generating the image does not take much power, though training them does; I should have been clearer.
When I say “massive amounts,” I mean a company like Microsoft opening large data centers that require enough energy and water to disrupt local communities. Obviously this isn’t an AI issue, and Microsoft doesn’t train for image generation AFAIK, but the fact remains that training an AI model requires an order of magnitude of more resources than most consumer or corporate applications.
If AI models were only getting more efficient, I wouldn’t worry about this, but companies tend to scale up and use more resources to make larger models.
Removed by mod