Suno, an AI music generator

AI generated content can be either derivative or transformative. For example if I use AI to paraphrase books or articles, that would be a derivative use.

But an AI that searches the web, news, and scientific papers for references and then outputs a Wikipedia-style article on a given topic would be a transformative use because it does a lot of work synthesizing multiple sources into a coherent piece, and only uses limited factual references.

Or we can do something more advanced. We solve a task with a student model and in parallel we solve the same task with a teacher model empowered with RAG, web search and code execution tools. Then you have two answers. You use an examination prompt to extract what the student model got wrong as a new training example.

That would be transformative, and targeted. No need to risk collecting content that is already known by the student model. It would be more like "machine studying" or "machine teaching" because it creates targeted lessons for the student.

It's going to Destroy sync no question. one of the few avenues artists have left to make an income with music. My personal opinion is that this is wholesale theft. While the technology is dazzling there is just no way suno wasn't trained on vast catalogues of copyrighted music.

What burns me up the most is that these people are walking around like they own the place. They don't. I really hope the NYT prevails in their copyright case. this will open a floodgate of lawsuits that will also be won.

On the creative side of generative AI, AI seems like a parasite that will do nothing more than further concentrate wealth into the hands of fewer and fewer people. When all of those robots come online and really start replacing jobs, there will be a reckoning and it's not going to be pretty.

> It's going to Destroy sync no question. one of the few avenues artists have left to make an income with music. My personal opinion is that this is wholesale theft. While the technology is dazzling there is just no way suno wasn't trained on vast catalogues of copyrighted music.

Yeah I've got a couple friends whose bread-and-butter has been sync for a couple decades, and they're really sweating. It's not just good for them, either-- their living off of that work gave them the money, time, and sophistication to hone their other music which benefits many people in many ways. They were even able to release a comedic side-project album after a particularly tragic school shooting for free, to national acclaim. Their being able to do that matters.

> What burns me up the most is that these people are walking around like they own the place. They don't. I really hope the NYT prevails in their copyright case. this will open a floodgate of lawsuits that will also be won.

Honestly, I left the software development business because that exact self-congratulatory unearned hubris that assumes expertise in everything because they understand how the experts' software was made. It's fucking ridiculous. The only people that think generative AI output is equivalent to making art have absolutely no idea how many decisions those algorithms make for the user, how incredibly consequential they are to the piece, and how much the original artists had to work to create them.

The most infuriating thing about this crowd is the arrogant, patronizing decrees many casually throw around about "the true" meaning of art and which artists are worthwhile and all that. They clearly have absolutely no clue what work artists even do in our society– most don't even have a functional definition of art, and couldn't make one piece of art scholarship that discusses it beyond whatever Aristotle they read in freshman Western History class. Their glib opinions are months old, largely adopted from other people in tech looking to justify not paying artists, and are based on some weird combination of fan art, fine art, and hobby art, being completely ignorant to the other 90% of the art world.

> On the creative side of generative AI, AI seems like a parasite that will do nothing more than further concentrate wealth into the hands of fewer and fewer people. When all of those robots come online and really start replacing jobs, there will be a reckoning and it's not going to be pretty.

Though I'm a commercial visual artist, and while there are analogs, I think there are important differences in both the market forces and disposition of generative AI in the practices and fields. I think there are some genuinely useful things that generative AI does in professional visual work-- it just doesn't look anything like what these products look like. In video compositing for example, masking foreground objects from background objects is a giant PITA, and a great use case for neural network algorithms. With Nuke's copycat functionality, you can mask an object in one frame in a video and it will replicate that (with varying success) to the rest of it. For 10 seconds of video, otherwise, that would be hundreds of images to individually mask, and it would be totally inconsistent. On a smaller scale is photoshop's foreground/background tools.

Market-wise, i think there's a larger variety of freelance jobs available in may art realms, but they're less reliable and pay worse-- even in graphic design. From the little experience I have in audio, compared to what I do, I think it takes more experience and skill to create a commercially usable lower-end audio deliverable and this can't do it yet, but the market can support far fewer people than the visual market(s) can so it's at even greater risk.

In visual art, lots of businesses are having generative AI make base assets and having their artists essentially clean them up to be professionally usable, which is about as soulless and shitty of a job as you can get. The end-user tools like Canva aren't much of a direct threat in the long-term to graphic designers because the use cases are usually a lot different, but my (less informed) gut says the real threat to commercial musicians is with products like Garage Band. I think there will be enough control to satisfy some art director's use case, and the output will eventually get high enough quality to hand a bunch of tracks to indie producers to "clean up" rather than making complete things, and that output will progressively get better and persistently decimate the market until only high-end specialists remain.