I guess I should start with my current views on AI-generated “art”. I think the motivations behind a lot of it are clear – to make money. Why should artists get paid to make the art I need for my project if I can just steal what they've already made, break it down along with the years of work of other artists, effectively laundering it. Then mash it all together into a paste that you can pour into a mold shaped like what you want without any of the effort. It was a rare wise decision by the American courts to deny the “artistic output” of any AI system to be uncopyrightable.
With that stated, and with some shame, I will admit that I've been having a lot of fun playing with the Udio AI-music generation website. (I feel weird calling it an app, despite the fact that there's a mobile version, but it's a website, the “app” runs on a website, I'm going to call it a website.) However, my reasons for enjoying it are most likely not what the people running it intended. I'm there to be subversive. I'm there to create mashups that would make Lucifer wince. I'm there to redefine musical cringe. I'm there to torture it into making the most amateurish, inept music imaginable. I'm there to make myself laugh.
When I visit the page, there is a section of songs which have been highlighted by staff – those songs have tens of thousands of views. Any other song? Maybe a half dozen to a dozen plays at best, and most of those are from the person who made it. Nobody is going to get famous at this. At best, it's a mechanized version of the old Song Poem scam – someone who fancies themselves a lyric writer would send their lyrics to a post-office box in Nashville, where the writer would almost certainly receive an acceptance letter, no matter how well-crafted or inept the lyrics actually were. They would be told for the sum of a certain amount of dollars, their song would be given the full “Nashville sound production”. Instead, a group of freelance session musicians would use their off-time performing on some big-name artists, to church out one-off improvisational recordings based on a stack of submitted, paid-for lyrics. They weren't always lyrics, sometimes they would be free-form poetry, occasionally gibberish. There was one case where a bored teenager, seeing the ad in a music magazine at the convenience market he worked at, realized it was a scam and set upon an idea. He would write the stupidest, most offensive lyrics he could think of, and send them in. Either he would get a rejection letter, which he could show to his friends with pride, or he would get a recording of his horrible lyrics. Either way, it was a win-win. Spoiler: He got his recording. You can find it on YouTube as “Peace and Love” by Ramsay Kearney (the session singer on the record). The song-poem industry died off a while, though, so AI will not be putting any song-poem companies out of business. In a way, I kind of hope services like Udio will help some people release the lyrical demon monkeys that have been flying around in their brains, and I'd love to bask in the insanity of the resulting 'music'.
But it will never replace the music creation process. Will there be AI music on the charts? Yeah, of course. Some people barely listen to music, they just have it on in the background. Those people won't care if it's human or AI. But they're not really music fans, per se.