81–90 of 283 posts
I really hope we move on from these boil-the-ocean models. I want something more collaborative and even iterative. I was having a conversation with a former bandmate. He was talking about a bunch of songs he is working on. He can play guitar, a bit of bass and can sing. That leaves drums. He wants a model where he can upload a demo and it either returns a stem for a drum track or just combines his demo with some drum…
I really hope we move on from these boil-the-ocean models. I want something more collaborative and even iterative. I was having a conversation with a former bandmate. He was talking about a bunch of songs he is working on. He can play guitar, a bit of bass and can sing. That leaves drums. He wants a model where he can upload a demo and it either returns a stem for a drum track or just combines his demo with some drum…
This is an incredible achievement, but as a musician, I wish this would go die in a fire. I'm a bedroom hobby musician with no dreams of ever making it big, but even so, I'm looking at the hours I'm spending trying to improve my skills and thinking what's the point, really, when I could just type in 'heavy metal guitar solo at 160bpm, A minor' and get something much much better? I know there is value in creating art…
When Harmonix was in the concept phase for Guitar Hero, there were two slides in the presentation. The first was the "con" slide: novel game style risks not finding a market, technical challenges of the new style of gameplay, requiring peripherals for a game is a famous pathway to low-volume sales. The next slide was labeled "pro," and it was just a picture of Jimi Hendrix on-stage mid-performance. I'd submit to you…
Earlier quoted context omitted.
I draw a very sharp line: curating the outputs and crafting the sentence is enough to make it art. If neither of those happen, it's just slop. It's actually a bit like photography. A bunch of randomly taken pictures piled together is not art. It needs to be done with purpose and refinement. Basically, in my own opinion, art ≠ a function of technical difficulty. Art = Curation, Refinement, and Taste
Disagree that curation and prompts adds artistry (dense intent reflected in the output) to AI generations. "Curation" in AI can only surface the curator's local maxima among a tiny and arbitrary grab-bag of seed integers they checked among the space of 2^64 options; it's statistically skewed 99% towards the model's whims rather than anyone's unique intent or taste. Prompt crafting is likewise terribly low fidelity si…
The models' latent space is extremely powerful, but you get hamstrung into the text encoders whims when you do things through a prompt interface. In particular, you've hit exactly an issue I have with current LLMs in general in that they are locked into wors and concepts that others have defined (labelings of points in the latent space).
Wishy washy thinking: I'd be nice if there were some sort of Turing complete lambda calculus sort of way to prompt these models instead. Where you can define new terms, create expressions, and loops and recursion or something.
It would sort of be like how SVGs are "intent complete" and undeniably art, but instead of vector graphics, it is an SVG like model prompt.
As a musician, that's what I find most compelling about Suno. It's become a tool to collaborate with, to help test out musical ideas and inspire new creativity. I listen to its output and then take the parts I like to weave into my own creations.
The AI music tools that generate whole songs out of prompts are a curious gimmick, but are sorely lacking in the above.
I really hope we move on from these boil-the-ocean models. I want something more collaborative and even iterative. I was having a conversation with a former bandmate. He was talking about a bunch of songs he is working on. He can play guitar, a bit of bass and can sing. That leaves drums. He wants a model where he can upload a demo and it either returns a stem for a drum track or just combines his demo with some drum…
Currently, all these AI tools generate the whole song which I'm not at all interested in given songwriting is so much fun
I really hope we move on from these boil-the-ocean models. I want something more collaborative and even iterative. I was having a conversation with a former bandmate. He was talking about a bunch of songs he is working on. He can play guitar, a bit of bass and can sing. That leaves drums. He wants a model where he can upload a demo and it either returns a stem for a drum track or just combines his demo with some drum…
I’d recommend GarageBand for this.
I believe there is massive room for improvement over what is currently available.
However, my larger point isn't "I want to do this one particular thing" and rather: I wish the music model companies would divert some attention away from "prompt a complete song in one shot" and towards "provide tools to iteratively improve songs in collaboration with a musician/producer".
Earlier quoted context omitted.
If that was true, then the development of a new urinal factory would have the same impact on art as the development of a new AI art models. The framing is dependent on the content
Framings require creators; they don't arise spontaneously. Someone could turn a urinal factory into art, but art doesn't validate itself. Belief alone in artistic essentialism doesn’t make it so.
The point I'm making is that a unique framing only results in a single piece of worthwhile conceptual art. You can't have an infinite factory of ducamp's fountain. What makes the piece worthwhile is that it was an original idea.
Conceptual art is different from decorational art in this sense. The AI music is a largely a homogenous synthesis of existing works. The AI "art" is decorational, not conceptual art. You could make an arrangement of AI art that is conceptual, but how many arrangements can you make that are actually worthwhile conceptually if AI art is generally homogenous?
It's like asking how many worthwhile works of conceptual art can you produce with a urinal factory that makes identical clones of the same urinal? 0 to 1.
And besides, it's not nessisarially true that all framings have creators. Nature is an example of a system that cultivates and curates a certain type of life without any rational process.