I have read Lanre Bakare’s article in The Guardian on how an AI-generated band amassed more than 1 million Spotify streams before admitting that they used AI to create their music, images, and backstory. The major concerns revolve around transparency, artist consent, and copyright infringement. Although I think artist consent and copyright infringement should be checked, I don’t find it necessary to introduce regulations for the streaming platforms to label AI-generated content just for the listeners to know the difference. Or what do you think?
I’m in the opposite camp; disclaimers about AI generation should be in place for music and video content. I believe it should be the consumer’s right to know if this content is AI-generated or not. Furthermore AI AI-generated content requires minimal creative effort and can be deployed at a faster rate than human content, which means that if there weren’t disclaimers, AI content could take over as people take advantage of it. It should be done for transparency purposes.
It quickly gets into dicey territory. What is considered free speech? And who has the intellectual property rights of what’s generated? I think that the space has plenty of opportunity to grow, but there is going to be some trouble as we get into the weeds with how AI intellectual property is claimed. To more directly address your point, depending on who “owns” the content may reframe the question to how it is regulated (by business, individual, model, etc).
I don’t think either of those questions you proposed are getting answered any time soon.
This is never really getting answered properly; this debate goes on forever. Because you then have questions such as “Is hate speech free speech?”, “Is deliberately lying about something free speech?”, “Is going against the government free speech?”, “Is going outside the status quo free speech?”, “Is calling for violence free speech?” and many more questions. No answer will satisfy everyone.
As for this, that is also very controversial, as I can see some error points:
- First, the prompter can get two different results from the same prompt list, so their input on the work is minimal, as seen by a previous message of mine.
- Meanwhile, the generative AI, as far as we are aware, is not sentient, and so cannot have creative input.
- The owners of the models could have broken copyright on training data as per a court case against Midjourney. BBC:Disney and Universal sue AI firm Midjourney over images. This makes it hard for them to have the claim
- The creators of the original content may or may not have contributed to the training data, and the AI per instance may or may not have resulted in copyrighted data outputted.
Realistically, it depends on a court case unless it gets settled out of court, then it’s still legal limbo. If Disney wins, the intellectual property is in the original creators, so using a model that generates Darth Vader will be breaking their copyright. If Midjourney wins, AI is either copyright-free, their copyright(so all AI-generated content belongs to the model owners, which they would like) or the user’s copyright(which makes conflicting prompts a bottomless pit of legal cases).
I think you provide a very insightful answer - and your retort is more or less why I asked the question. It gets incredibly slippery to define what is whose in the great game of AI generation, and I think that as AI becomes more proficient, it’ll only get more difficult. Furthermore, it’ll get even more dicey as different countries enforce different laws or violate the laws of other countries.
I agree. There are two primary ways I look at it.
- The standpoint of embracing AI. Even if AI will be able to avoid stealing styles in the future, it doesn’t seem to be there yet. I heard some songs from this band, and they reeked of songs from generation you-know-who, but maybe didn’t break any laws. In other cases, AI outright copies licks. https://youtu.be/RNHxHFlGSkI?si=rODlOQwwPQYK4-39
It’s all fun and games up to a point, and I lean toward anything that genuinely allows AI growth, but only up to a point.
I know what it’s like to be ripped off. I’ve seem my literature published as someone else’s work. I didn’t catch the name of the title, but it was some of my best and a lot more than the legal amount one can quote, plus, I wasn’t credited. Maybe it was AI, maybe not, but you get my point. And now there’s nothing I can do because I didn’t catch the title.
That pain point should be avoided as much as possible. Believe me. It hurts.
- A legal point of view. People have been covering each other’s music for ages. There are laws in place that govern that. Maybe those laws will carry over, or maybe they need to be changed, but AI probably needs to be governed so that artists are protected.
As you said, it gets into dicy territory. Here’s an example: Someone spends an hour to prompt a song. Who owns rights to that song, AI, or the prompter?