5 ms·

I don't misunderstand it whatsoever, you just choose to ignore that consent trumps any notion of commercialization and doing as you please, in both a legal and

I don't misunderstand it whatsoever, you just choose to ignore that consent trumps any notion of commercialization and doing as you please, in both a legal and ethical context. If these things don't need to be trained on the work of others for whom they don't have consent and have not paid a license fee to, then why are they? Because it's exploitation by some of the biggest corps in the world. They should not be above the law and the utmost basic of ethics—consent. That is why it isn't "okay". I am not debating that generative AI can't be done ethically, I am pointing out that search is consensual. I can use robots.txt to opt out of being included in search results, but search engines direct traffic to the author. Authors did not consent to having their non-web content included in an LLM as an example. The problem is consent. The problem is your attitude towards consent. The primary purpose of the commercially and publicly available generative AI is to undercut human labour. Fact. What it serves up is frequently too similar to the existing work that it was trained on, and therein lies the problem on every level, not just from copyright infringement, but one of consent. You approached it from the angle of copyright, but the problem is one of consent. The reason it isn't "okay" is because consent wasn't given with these public, commercialized models. Deepfakes are part-and-parcel to copyright infringement. If you can take x and make it look more like y by applying z attributes then this is why you can make it regurgitate someone's intellectual property who did not consent to being included.

golol 3y ago
> If these things don't need to be trained on the work of others for whom they don't have consent and have not paid a license fee to, then why are they? You made a point about meta training on non-publicly viewable/freely available works, such as pirated movies. That's why I made the point that using non-publicly viewable data is sufficient. At the same time I said that, using data for which you have neither consent nor license is indeed necessary, and for the training on such data the copyright/fair use considerations are complicated. > I am pointing out that search is consensual. If I train a diffusion model on images from google search,and make a paid latent-space search where I just show people some pictures without any claims of selling them rights to anything, then I claim that this is just as consensual as google images. That is because I took the source images from google images where they were consensually posted, and in the worst case my model is exactly reproducing them. More likely my model is only approximately/interpolatively reproducing them. Sure, you might change your robots.txt to update google image behavior and while model won't update, but I think generally if your images has been reposted, it is quite likely that your image is still searchable on google images, even if you change the robots.txt for your personal blog. So it's not like google is very good at updating consent either. Perhaps there is a point to be made that my diffusion-search may not just reproduce your works approximately, for which I claim that consent is kind of given, but furthermore reworks or recombines them with other works. There it is perhaps true that new consent should be necessary.
golol 3y ago
> The primary purpose of the commercially and publicly available generative AI is to undercut human labour. As is the primary purpose of almost all technology, so not per se a bad thing. > What it serves up is frequently too similar to the existing work that it was trained on, and therein lies the problem on every level, not just from copyright infringement, but one of consent. What's the difference to google images then? If the problem is only that people get to see something similar to my work, then google iamges is just as bad. Well perhaps the difference for you is consent, in which case see above. But for me the difference was that OpenAI is trying to sell you rights for thise things that look similar to your works. So for me, if OpenAI just stopped trying to sell you rights for its generations, it would be copyright-okay. > You approached it from the angle of copyright, but the problem is one of consent. Fair enough, that is true. And consent is also an interesting topic to discuss. For me personally, with consent issues one has to think wheter one is talking about being ethical, or nice, or talking about legal action being necessary. I personally think that the consent issues with genAI are more in the "soft" side, compared to the copyright issues. With the latter it is a very pressing issue to find out how it should be treated legally.

golol 3y ago

> If these things don't need to be trained on the work of others for whom they don't have consent and have not paid a license fee to, then why are they? You made a point about meta training on non-publicly viewable/freely available works, such as pirated movies. That's why I made the point that using non-publicly viewable data is sufficient. At the same time I said that, using data for which you have neither consent nor license is indeed necessary, and for the training on such data the copyright/fair use considerations are complicated. > I am pointing out that search is consensual. If I train a diffusion model on images from google search,and make a paid latent-space search where I just show people some pictures without any claims of selling them rights to anything, then I claim that this is just as consensual as google images. That is because I took the source images from google images where they were consensually posted, and in the worst case my model is exactly reproducing them. More likely my model is only approximately/interpolatively reproducing them. Sure, you might change your robots.txt to update google image behavior and while model won't update, but I think generally if your images has been reposted, it is quite likely that your image is still searchable on google images, even if you change the robots.txt for your personal blog. So it's not like google is very good at updating consent either. Perhaps there is a point to be made that my diffusion-search may not just reproduce your works approximately, for which I claim that consent is kind of given, but furthermore reworks or recombines them with other works. There it is perhaps true that new consent should be necessary.

golol 3y ago

> The primary purpose of the commercially and publicly available generative AI is to undercut human labour. As is the primary purpose of almost all technology, so not per se a bad thing. > What it serves up is frequently too similar to the existing work that it was trained on, and therein lies the problem on every level, not just from copyright infringement, but one of consent. What's the difference to google images then? If the problem is only that people get to see something similar to my work, then google iamges is just as bad. Well perhaps the difference for you is consent, in which case see above. But for me the difference was that OpenAI is trying to sell you rights for thise things that look similar to your works. So for me, if OpenAI just stopped trying to sell you rights for its generations, it would be copyright-okay. > You approached it from the angle of copyright, but the problem is one of consent. Fair enough, that is true. And consent is also an interesting topic to discuss. For me personally, with consent issues one has to think wheter one is talking about being ethical, or nice, or talking about legal action being necessary. I personally think that the consent issues with genAI are more in the "soft" side, compared to the copyright issues. With the latter it is a very pressing issue to find out how it should be treated legally.