• 0 Posts
  • 11 Comments
Joined 1 year ago
cake
Cake day: June 16th, 2023

help-circle


  • There are many different niches of ML. 99% of hobbyist would use consumer grade hardware. It’s quite frankly more than good enough.

    Even in commercial usage, consumer GPUs provide better value unless you need to do something that very specifically require a huge vram pool. Like connecting multiple A100 GPUs to have hundreds or tens of thousands of gigabyte vram. Those use cases only come up if you’re making base models for general purpose.

    If you’re using it for single person use case, something like 4090 is actually the best hardware. Enough ram to run almost anything and it’s higher clock speed than enterprise GPU means your results come back faster.

    Even training doesn’t require that much vram. Chat models are generally more vram heavy but if you’re doing specific image training like stable diffusion for how to render your face, or some specific fetish porn, you only really need like 12GB of vram to do it. There are ways to even do it at lower like 8GB but 12 is sweet value spot where even 3060 or 4060ti can do. Consumer GPUs will get that trained in like 30min to 24hrs depending on settings and model.



  • Grumpy@sh.itjust.workstoGames@sh.itjust.worksCD Projekt employees form a union
    link
    fedilink
    English
    arrow-up
    0
    arrow-down
    1
    ·
    edit-2
    9 months ago

    I’m all for unions. But I’m not sure how it translates to good for players. Unions exist for fair wages and working environment, not direction of how games should be made.


    Edit: People sure seem to get the wrong impression with my question. As I said in the very first line, I am for unions. They’re great and we should strive for fair working wages and hours, especially in 2023 where wages are stagnating while having massive inflation. We should have happy employees and I prefer my games made by happy employees. Failure to keep the wages up is creating shit ton of societal problems.

    Issue is the delusion people are presenting here. Unions are not magic. It doesn’t automatically improve unrelated things. What people are missing is that there is no evidence the union has ever advocated for a better product. If one exists, despite my desperate attempt to find one, then it’s clearly a fringe case. All the replies are making a huge logical leap of simply saying happy worker produces better product with no reasoning behind it. Unions never argue for better product. That’s just not what unions do. It argues for the betterment of workers.

    Unionizing increases productivity for some sectors. But they’re usually rare and only seen in specific industries. They generally have no significant impact on productivity based on research. If it straight up increased productivity and made better products, every company would love it. The argument is counter-logical. Companies do what is efficient. Even if we assumed individual productivity is increased, there’s still no evidence that these individuals would have the capacity to change the direction in which the product is being made in the upper tier.

    We need unions. But unions aren’t magic.




  • Sorry, this is wrong.

    As a general statement: No, I am not. You’re making an over specific scenario to make it true. Sure, if I take 1 image and train a model just on that one image, it’ll make that exact same image. But that’s no different than me just pressing copy and paste on a single image file. The latter does the job whole lot better too. This entire counter argument is nothing more than being pedantic.

    Furthermore, if I’m making such specific instructions to the AI, then I am the one who’s replicating the art. It doesn’t matter if I use a pencil to trace out the existing art, using photoshop, or creating a specific AI model. I am the one who’s doing that.


  • What gives a human right to learn off of another person without credit? There is no such inherent right.

    Even if such a right existed, I as a person who can make AI training, would then have the right to create a tool to assist me in learning, because I’m a person with same rights as anyone else. If it’s just a tool, which it is, then it is not the AI which has the right to learn, I have the right to learn, which I used to make the tool.

    I can use photoshop to replicate art a lot more easily than with AI. None of us are going around saying Photoshop is wrong. (Though we did say that before) The AI won’t know any specific art unless it’s an extremely repeated pattern like “mona lisa”. It literally do not have the capacity to contain other people’s art, and therefore it cannot replicate others art. I have already proven that mathematically.



  • If AI art is stolen data, then every artists on earth are thieves too.

    Do you think artists just spontaneously conjure up art? No. Through their entire life of looking at other people’s works, they learned how to do stuff, they emulate and they improve. That’s how human artists come to be. Do you think artists go around asking permission from millions of past artists if they can learn from their art? Do artists track down whoever made the fediverse logo if I want to make a similar shaped art with it? Hell no. Consent in general is impossible too because whole lot of them are likely too dead to give consent be honest. Its the exact same way AI is made.

    Your argument holds no consistent logic.

    Furthermore, you likely have a misunderstanding of how AI is trained and works. AI models do not store nor copy art that it’s trained on. It studies shapes, concepts, styles, etc. It puts these concepts into matrix of vectors. Billions of images and words are turned into mere 2 gigabytes in something like SD fp16. 2GB is virtually nothing. There’s no compression capable of anywhere near that. So unless you actually took very few images and made a 2GB model, it has no capability to store or copy another person’s art. It has no knowledge of any existing copyrighted work anymore. It only knows the concepts and these concepts like a circle, square, etc. are not copyrightable.

    If you think I’m just being pro-AI for the sake of it. Well, it doesn’t matter. Because copyright offices all over the world have started releasing their views on AI art. And it’s unanimously in agreement that it’s not stolen. Furthermore, resulting AI artworks can be copyrighted (lot more complexity there, but that’s for another day).