Yeah, Lemmy is a bit over-the-top anti-AI, but most of it is based in reality.
There are a bunch of problems with AI. And they outbnumer any good ones by a mile.
The main cause of that fact is the entire AI bubble.
AI wastes a fuckton of energy. Of course, this energy isn’t free: communities pay. Electricity demand goes up, and so does price. Then, most electricity isn’t green. And on top of that, the rise in demand causes more electricity peaks, which almost exclusively get “fixed” through fossil fuel-based methods.
From another angle, AI disrupts markets. And not in a good way. Companies dump millions into AI while neglecting their employees (who get laid off because AI “can replace” them), and their customers as well (since instead of doing useful stuff for consumers they pump out AI-branded bullshit no one wants or needs).
Then, big AI companies spit in the face of copyright and have the audacity to turn around and claim copyright on their models’ outputs. If inputs are free game, so are the outputs. Copyright is a very vague, misunderstood and misused term, and no argument I’ve heard claiming feeding stuff into AI is fair use was grounded in reality.
That all veing said, AI is here to stay. I’ve been thinking long and hard about similar fundamental changes to how human society functions, and I think i found one. Photography.
Way back when, you had to do things painstakibgly by hand. Drawing, copying books by hand, etc.
Then the printing press came. Revolutionary? Sure. But not as revolutionary as photography. Instead of writing by hand, you had to typeset by hand before printing. This made the process scalable, but it was still painstaking work.
But photography is a different matter. You just have to make (or buy) a camera and other required supplies (film, developing media, etc), and then you merely have to set up the camera, take the photo, develop the film, and make the photo.
Even in the early days of photography, while these processes took some time, it wasn’t painstaking. To take a photo, you set up the camera, and wait. To develop film, you dunk the film into a chemical bath, and wait. To transfer the image onto paper - a similar ordeal. Set, forget.
Photography fundamentally changed how the entirety of society works. Painters complained and lost jobs and livelihoods - like the “jobs stolen” by AI. Instead of drawing stuff, which required a lot of skill, taking a photo is much simpler (abd faster).
Yesterday, instead of having to paint stuff, you’d take a photo. Today, instead of taking a photo, you ask AI.
On the copyright front, the paralels are obvious: Taking a photo of a book is fair use. But photocopying a book isn’t. The problem with AI is that it does some transformations to the original, so it’s obfuscated inside the model. But the obfuscation can be undone, as AI often happily spits out certain inputs verbatim when asked. Take a photo of a page - okay. Photocopy the entire book? Not okay.
The situation is the same when we look at artwork instead of books. Taking a photo of an artwork in a museum is okay. Scanning an artwork (duplicating it verbatim) - isn’t. Same for movies. A frame is probably gonna be okay. The entire movie - won’t.
Going by the closest analogue, there is absolutely no justification to indiscriminately feed everything and anything into AI, for indiscriminately photocopying and vervatim copying the same material is clearly protected.
Yeah, Lemmy is a bit over-the-top anti-AI, but most of it is based in reality.
There are a bunch of problems with AI. And they outbnumer any good ones by a mile.
The main cause of that fact is the entire AI bubble.
AI wastes a fuckton of energy. Of course, this energy isn’t free: communities pay. Electricity demand goes up, and so does price. Then, most electricity isn’t green. And on top of that, the rise in demand causes more electricity peaks, which almost exclusively get “fixed” through fossil fuel-based methods.
From another angle, AI disrupts markets. And not in a good way. Companies dump millions into AI while neglecting their employees (who get laid off because AI “can replace” them), and their customers as well (since instead of doing useful stuff for consumers they pump out AI-branded bullshit no one wants or needs).
Then, big AI companies spit in the face of copyright and have the audacity to turn around and claim copyright on their models’ outputs. If inputs are free game, so are the outputs. Copyright is a very vague, misunderstood and misused term, and no argument I’ve heard claiming feeding stuff into AI is fair use was grounded in reality.
That all veing said, AI is here to stay. I’ve been thinking long and hard about similar fundamental changes to how human society functions, and I think i found one. Photography.
Way back when, you had to do things painstakibgly by hand. Drawing, copying books by hand, etc.
Then the printing press came. Revolutionary? Sure. But not as revolutionary as photography. Instead of writing by hand, you had to typeset by hand before printing. This made the process scalable, but it was still painstaking work.
But photography is a different matter. You just have to make (or buy) a camera and other required supplies (film, developing media, etc), and then you merely have to set up the camera, take the photo, develop the film, and make the photo.
Even in the early days of photography, while these processes took some time, it wasn’t painstaking. To take a photo, you set up the camera, and wait. To develop film, you dunk the film into a chemical bath, and wait. To transfer the image onto paper - a similar ordeal. Set, forget.
Photography fundamentally changed how the entirety of society works. Painters complained and lost jobs and livelihoods - like the “jobs stolen” by AI. Instead of drawing stuff, which required a lot of skill, taking a photo is much simpler (abd faster).
Yesterday, instead of having to paint stuff, you’d take a photo. Today, instead of taking a photo, you ask AI.
On the copyright front, the paralels are obvious: Taking a photo of a book is fair use. But photocopying a book isn’t. The problem with AI is that it does some transformations to the original, so it’s obfuscated inside the model. But the obfuscation can be undone, as AI often happily spits out certain inputs verbatim when asked. Take a photo of a page - okay. Photocopy the entire book? Not okay.
The situation is the same when we look at artwork instead of books. Taking a photo of an artwork in a museum is okay. Scanning an artwork (duplicating it verbatim) - isn’t. Same for movies. A frame is probably gonna be okay. The entire movie - won’t.
Going by the closest analogue, there is absolutely no justification to indiscriminately feed everything and anything into AI, for indiscriminately photocopying and vervatim copying the same material is clearly protected.