4 ms·

My view is that if AI is trained on public data and it doesn't need to pay royalties for it, then it doesn't make sense that the public should pay royalties to

My view is that if AI is trained on public data and it doesn't need to pay royalties for it, then it doesn't make sense that the public should pay royalties to use the AI's output. If AI companies paid royalties for their entire training set, then it would be a different situation. But given that it may be infeasible to prove or disprove whether or not AI was trained on public data, we should err on the side of caution and not allow the companies which build/operate AI to have copyrights over its output. A case could be made that most work which humans produce is itself derivative of various public works and therefore the AI is not so different from humans and therefore it should be given copyrights... But I think the key difference is that the AI is not human, there are 'human rights' but there are no 'AI rights' or 'machine rights'... Until an AI can prove that it has agency and that it works for its own interests and not for a third party, it should not have any rights.

tanseydavid 3y ago
You seem to only be considering IP costs. To me, it seems that the bigger, on-going issue, is "Who will pay for the compute charges?" (which are not trivial)
- jongjong 3y ago
  I feel like it's companies' problem if their business model is not compatible with the new reality of AIs relentlessly scraping their data. They may need to rethink their business model and try to increase their margins and/or improve their efficiency to account for rising hosting costs. I guess there are difficult times ahead. I think it's not so bad since the entities who train and/or use AI also have to pay a lot of money on their end.
Auracle 3y ago
Alright - what if I take Stable Diffusion and further train it on my own art, and then use that to generate things? Or what if I don’t do that, but generate several images and spend hours making a composite? What if I start with my own photograph and generate things to use in a composite on top of that? It’s not so clear cut.
- CRConrad 3y ago
  > Alright - what if I take Stable Diffusion and further train it on my own art, and then use that to generate things? Operative word there: "further". If it's already trained on everyone else's input, then without that there wouldn't even be a Stable Diffusion for you to "further train on your own art", so YTF should you get sole ownership of the sum of it all? What you're saying is like "Alright - what if I take Linux and add in my own code, and then sell that as closed source?" Nope, it's already GPLed, and so should AIs trained on public data be.
- jongjong 3y ago
  It is clear cut because the Stable Diffusion AI you refer to was nevertheless trained on public data to get to that point were it could then be 'fine-tuned' with your own art. Your own art is just a small addition to its much, much larger input/training set. The size of these public training sets are truly massive. If you start with a Stable Diffusion program from scratch with all weights initialized at random (or whatever weights its initialization algorithm requires) and you train it on only your own photograph(s), then you will almost certainly find that it's not effective at all. In that case though, the low-quality output would be 100% your own. The reason why generative AIs are so good is precisely because they have access to a large body of public data. Without the public data, they would not be good no matter how good the underlying software/algorithm is.
YetAnotherNick 3y ago
No, I would hate this regulation and this is everything against what open internet stood for. I can see that you don't want AI companies to exist so you want this regulation, but how is different than search engine scraping the internet? If the content is not behind TOS or paywall, it should be equally scrapable by anyone be it search engine or AI companies or me. If the content is copyrighted, it should only be shown if user accepts the terms or pays for it.
- daveguy 3y ago
  > I can see that you don't want AI companies to exist so you want this regulation, but how is different than search engine scraping the internet? This is unnecessary hyperbole. Preventing the laundering of copyright material through AI algorithms does not prevent AI companies from existing. Take Adobe -- they have license to the training images they use and they have helpful products based on AI processing of those images. It won't be "free" money from dubious sources. But AI does not depend on the laundering of copyright material. The difference is whether or not you are scraping and then generating similar products (eg the images themselves) or whether you are scraping for some other purpose.
- jongjong 3y ago
  I believe there are rules relating to copyright which allow using up to a certain number of words of text from an original text without infringing on copyright. I think a difference with search engines there is that search engines reference the source with a link whereas the AI does not. Also, critically, search engines do not claim copyright over the text shown in their results pages. I think that forms the legal basis for why AIs should be allowed to train on public data for free but it does not form a legal basis for AI to have copyrights over the output.

tanseydavid 3y ago

You seem to only be considering IP costs. To me, it seems that the bigger, on-going issue, is "Who will pay for the compute charges?" (which are not trivial)

jongjong 3y ago
I feel like it's companies' problem if their business model is not compatible with the new reality of AIs relentlessly scraping their data. They may need to rethink their business model and try to increase their margins and/or improve their efficiency to account for rising hosting costs. I guess there are difficult times ahead. I think it's not so bad since the entities who train and/or use AI also have to pay a lot of money on their end.

Auracle 3y ago

Alright - what if I take Stable Diffusion and further train it on my own art, and then use that to generate things? Or what if I don’t do that, but generate several images and spend hours making a composite? What if I start with my own photograph and generate things to use in a composite on top of that? It’s not so clear cut.

CRConrad 3y ago
> Alright - what if I take Stable Diffusion and further train it on my own art, and then use that to generate things? Operative word there: "further". If it's already trained on everyone else's input, then without that there wouldn't even be a Stable Diffusion for you to "further train on your own art", so YTF should you get sole ownership of the sum of it all? What you're saying is like "Alright - what if I take Linux and add in my own code, and then sell that as closed source?" Nope, it's already GPLed, and so should AIs trained on public data be.
jongjong 3y ago
It is clear cut because the Stable Diffusion AI you refer to was nevertheless trained on public data to get to that point were it could then be 'fine-tuned' with your own art. Your own art is just a small addition to its much, much larger input/training set. The size of these public training sets are truly massive. If you start with a Stable Diffusion program from scratch with all weights initialized at random (or whatever weights its initialization algorithm requires) and you train it on only your own photograph(s), then you will almost certainly find that it's not effective at all. In that case though, the low-quality output would be 100% your own. The reason why generative AIs are so good is precisely because they have access to a large body of public data. Without the public data, they would not be good no matter how good the underlying software/algorithm is.

YetAnotherNick 3y ago

No, I would hate this regulation and this is everything against what open internet stood for. I can see that you don't want AI companies to exist so you want this regulation, but how is different than search engine scraping the internet? If the content is not behind TOS or paywall, it should be equally scrapable by anyone be it search engine or AI companies or me. If the content is copyrighted, it should only be shown if user accepts the terms or pays for it.

daveguy 3y ago
> I can see that you don't want AI companies to exist so you want this regulation, but how is different than search engine scraping the internet? This is unnecessary hyperbole. Preventing the laundering of copyright material through AI algorithms does not prevent AI companies from existing. Take Adobe -- they have license to the training images they use and they have helpful products based on AI processing of those images. It won't be "free" money from dubious sources. But AI does not depend on the laundering of copyright material. The difference is whether or not you are scraping and then generating similar products (eg the images themselves) or whether you are scraping for some other purpose.
jongjong 3y ago
I believe there are rules relating to copyright which allow using up to a certain number of words of text from an original text without infringing on copyright. I think a difference with search engines there is that search engines reference the source with a link whereas the AI does not. Also, critically, search engines do not claim copyright over the text shown in their results pages. I think that forms the legal basis for why AIs should be allowed to train on public data for free but it does not form a legal basis for AI to have copyrights over the output.