The question here then becomes whether AI can actually be creative, or if it's simply producing remixed versions of its influences according to a set of instructions in the programmer's algorithm, along with instructions ordering production of a given piece. The AI's database establishes exposure to the copyrighted works. If the AI algorithm simply remixes and distorts its influences, there's no creativity. It's simply following instructions telling it how much of which influences to use in the remixed end-product. It's nothing more than the sum of its influences. To avoid the infringement case, it must be demonstrated that the AI introduces something independently creative.
I think you are making a mistake of separating the so-called ‘AI’ from the persons who programmed it. The litigants, in this case, are suing the programmers or owners of the technology, as they are subject to the law. AIs, like self-driving cars, are merely technologies. They only do what they are programmed to do. We should be careful not to assume that AIs have agency.
Fascinating topic and discussion. I think the litigants have a case, but I’m no law expert. This reminds me of the US courts ruling that a computer loading data in to RAM is technically making a ‘copy,’ and that must be expressly permitted by the copyright holder, or else it is a violation of their copyright. http://digital-law-online.info/lpdi1.0/treatise20.html I don’t know whether subsequent court rulings have superseded this interpretation of the law. Maybe artists in future will have to expressly permit or forbid AI from ‘scraping’ their work, and stuff in the public domain will remain fair game.
I meant to say that if the AI, in the process of ‘learning,’ brings an artwork into RAM — thatis a copy, and therefore copyright has potentially been infringed, according to a previous ruling of the courts… as I understand it.
I think you are making a mistake of separating the so-called ‘AI’ from the persons who programmed it. The litigants, in this case, are suing the programmers or owners of the technology, as they are subject to the law. AIs, like self-driving cars, are merely technologies. They only do what they are programmed to do. We should be careful not to assume that AIs have agency.
I think I was not making that mistake.
Fair enough. I misunderstood your comment.
I’m struggling with the thought that an AI might be creative, which is apparently what their programmers are trying to achieve — the appearance that this thing can creatively paint, or write, or compose music etc., to a certain level of competence. I presume that the painting AIs are merely raiding their store of images and associated keywords, derived from scraping art works by real human artists. Which is why most AI art looks clichéd to me, e.g. putting a zombie in a picture of ‘the end of the world.’
- archive.org caches copyrighted images aplenty - google and other search engines do the same - your browser, same
There's something to be said about Terms of Service that allows for specific use cases where an automated system is allowed to access resources for specific purposes and with specific limitations (that is, search engines can't display the full text of articles - only a snippet, etc.). If the AI models were trained on images via a process that violated ToS, then yes, they'd be in violation of those terms, and probably copyright as well (IANAL). However, I'm not sure there's proof the plaintiffs can bring that their specific works were accessed in violation of ToS during the training process. If they can, they might have a leg to stand on and the models might need to be retrained a bit more carefully, but could still perform similarly.
I'm guessing not every published resource on the web must have a ToS explicitly enabling machine bulk access like spidering/crawling, or search engines could never exist. I therefore assume only websites that publish a ToS with language explicitly restricting machine access could be considered protected from such.
On the question of derivation, if a human can look at images, learn from them, and then have in mind the ideas of what a cat or a car looks like, or develop a heuristic for what a given artist's style is, then I think using a tool to do the same (look at images, learn from them, and generate something based on that learning) is no different. Unless I greatly misunderstand what the model contains or how it works, it doesn't retain a copy of all the images, nor does it reference them during generation - only it's "learning". I'm fine with that, even if it upends some things.
Comments
- archive.org caches copyrighted images aplenty
- google and other search engines do the same
- your browser, same
There's something to be said about Terms of Service that allows for specific use cases where an automated system is allowed to access resources for specific purposes and with specific limitations (that is, search engines can't display the full text of articles - only a snippet, etc.). If the AI models were trained on images via a process that violated ToS, then yes, they'd be in violation of those terms, and probably copyright as well (IANAL). However, I'm not sure there's proof the plaintiffs can bring that their specific works were accessed in violation of ToS during the training process. If they can, they might have a leg to stand on and the models might need to be retrained a bit more carefully, but could still perform similarly.
I'm guessing not every published resource on the web must have a ToS explicitly enabling machine bulk access like spidering/crawling, or search engines could never exist. I therefore assume only websites that publish a ToS with language explicitly restricting machine access could be considered protected from such.
On the question of derivation, if a human can look at images, learn from them, and then have in mind the ideas of what a cat or a car looks like, or develop a heuristic for what a given artist's style is, then I think using a tool to do the same (look at images, learn from them, and generate something based on that learning) is no different. Unless I greatly misunderstand what the model contains or how it works, it doesn't retain a copy of all the images, nor does it reference them during generation - only it's "learning". I'm fine with that, even if it upends some things.