More like I can’t sell photographs of turnips if I have to pay to take photos of them. Why should we have to pay to take photos of turnips when we never have had to ever?
They’re someone else’s turnips though, not yours. If you’re going to make money selling pictures of them, don’t you think the person who grew the turnips deserves a fair share of the proceeds?
Or from another perspective, if the person who grew them requests payment in return for you to take pictures of them, and you don’t want to pay it – why don’t you go find other turnips? Or grow your own?
These LLMs are an end product of capitalism – exploiting other people’s labor and creativity without paying them so you can get rich.
To answer your first question: No I don’t think the person growing turnip that I can see from the street should be compensated for the photograph I sell of that turnip. What next ? should we also compensate his parents for teaching him how to grow turnip, or his grandparent for teaching his parents ? What about the architect who designed the house next door that you can see in the background of the photograph ? Should the maker of the camera be compensated every time I take a picture ?..
Anyway back to AI:
I think though that the AI model resulting from freely accessing all images should also be fully open source and that anyone should be allowed to locally execute it on their own hardware. Let’s use this to push for the end of Intellectual property.
That’s a slippery slope fallacy. We can compensate the person with direct ownership without going through a chain of causality. We already do this when we buy goods and services.
I think the key thing in what you’re saying about AI is “fully open source… locally execute it on their own hardware”. Because if that’s the case, I actually don’t have any issues with how it uses IP or copyright. If it’s an open source and free to use model without any strings attached, I’m all for it using copyrighted material and ignoring IP restrictions.
My issue is with how OpenAI and other companies do it. If you’re going to sell a trained proprietary model, you don’t get to ignore copyright. That model only exists because it used the labor and creativity of other people – if the model is going to be sold, the people whose efforts went into it should get adequately compensated.
In the end, what will generative AI be – a free, open source tool, or a paid corporate product? That determines how copyrighted training material should be treated. Free and open source, it’s like a library. It’s a boon to the public. But paid and corporate, it’s just making undeserved money.
Funny enough, I think when we’re aligned on the nature and monetization of the AI model, we’re in agreement on copyright. Taking a picture of my turnips for yourself, or to create a larger creative project you sell? Sure. Taking a picture of my turnips to use in a corporation to churn out a product and charge for it? Give me my damn share.
