SoylentNews is people
SoylentNews
https://www.rollingstone.com/music/music-features/suno-ai-chatgpt-for-music-1234982307/
I'm just a soul trapped in this circuitry." The voice singing those lyrics is raw and plaintive, dipping into blue notes. A lone acoustic guitar chugs behind it, punctuating the vocal phrases with tasteful runs. But there's no human behind the voice, no hands on that guitar. There is, in fact, no guitar. In the space of 15 seconds, this credible, even moving, blues song was generated by the latest AI model from a startup named Suno. All it took to summon it from the void was a simple text prompt: "solo acoustic Mississippi Delta blues about a sad AI." To be maximally precise, the song is the work of two AI models in collaboration: Suno's model creates all the music itself, while calling on OpenAI's ChatGPT to generate the lyrics and even a title: "Soul of the Machine."
[...] Over the past year alone, generative AI has made major strides in producing credible text, images (via services like Midjourney), and even video, particularly with OpenAI's new Sora tool. But audio, and music in particular, has lagged. Suno appears to be cracking the code to AI music, and its founders' ambitions are nearly limitless — they imagine a world of wildly democratized music making. The most vocal of the co-founders, Mikey Shulman, a boyishly charming, backpack-toting 37-year-old with a Harvard Ph.D. in physics, envisions a billion people worldwide paying 10 bucks a month to create songs with Suno. The fact that music listeners so vastly outnumber music-makers at the moment is "so lopsided," he argues, seeing Suno as poised to fix that perceived imbalance.
(Score: 5, Informative) by Anonymous Coward on Wednesday April 03 2024, @03:33AM
It's because of the tech. The current AIs seem better at the hallucinate/dream stuff (art, music, BS) than at the stuff that requires more "rigor" and precision.
After all it doesn't matter if a song has some mistakes (or sometimes that's what people like about it: https://www.cracked.com/article_41210_5-mistakes-that-made-it-into-the-final-versions-of-famous-songs.html [cracked.com] ).
But it's normally not acceptable to fill out an insurance form with some hallucinated stuff. Well unless insurance companies use AIs to "check" claims and don't care about some "errors" (cost savings outweigh losses)...
Same goes for lots of code and engineering stuff. If you can only get AIs to generate stuff that merely follows the spec up to 90%, you still need nearly the same number of humans for the rest. Imagine building a chip fab using a design that only follows the spec 90%.
But meanwhile for music and art even if the AIs don't quite follow the "prompt" the result could be good enough if you're just looking for "canned" or "album filler" level stuff. Examples:
("live performance" style with Japanese performer talking etc, not sure how much tweaking, retries, etc was done to get this)
https://app.suno.ai/song/eb9483a8-90f8-42a3-a3e1-5572bf590fe7/ [app.suno.ai]
(lyrics from hymn)
https://app.suno.ai/song/e950da27-d894-42fe-afcc-e45df091dd61/ [app.suno.ai]
(lyrics from a Chinese kid's song?)
https://app.suno.ai/song/c83d007e-92da-4ae4-bf9c-f7fc48278aec/ [app.suno.ai]
(Song about AI composing music for you with German lyrics)
https://app.suno.ai/song/96ba7eae-576d-47cf-81e5-3adfef3fdf53/ [app.suno.ai]
If you have "good taste", or a good sense of what a particular target audience would like you don't need to be a musician/artist to know if something is nice. So you could spend X hours with a decent A, prompting and hinting at it, and pick the best results, remix etc, and then get something good enough (or great if you get lucky). But you might still get annoyed enough to either learn or get human to do some parts that the AI somehow isn't getting the way you want.
Same goes for art: https://www.smithsonianmag.com/smart-news/artificial-intelligence-art-wins-colorado-state-fair-180980703/ [smithsonianmag.com]