Beatles' 'Now and Then' Makes History As First AI-Assisted Song To Earn Grammy Nomination (billboard.com)
- Reference: 0175452057
- News link: https://entertainment.slashdot.org/story/24/11/11/2132242/beatles-now-and-then-makes-history-as-first-ai-assisted-song-to-earn-grammy-nomination
- Source link: https://www.billboard.com/pro/the-beatles-now-and-then-first-ai-assisted-song-grammy-nomination-history
> When "Now and Then" first came out in late 2023, the disclosure that it was finalized utilizing AI caused an uproar. At the time, many fans assumed that the remaining Fab Four members -- Paul McCartney and Ringo Starr -- must have used generative AI to deepfake the late John Lennon. That was not actually the case. Instead, the Beatles used a form of AI known as "stem separation" to help them clean up a 60-year-old, low-fidelity demo recorded by Lennon during his lifetime and to make it useable in a finished master recording.
>
> With stem separation, the Beatles could isolate Lennon's vocal and get rid of excess noise. Proponents of this form of technology say it has major benefits for remastering and cleaning up older catalogs. Recently, AudioShake, a leading company in this space, struck a partnership with Disney Music Group to help the media giant clean up its older catalog to "unlock new listening and fan engagement experiences" like lyric videos, film/TV licensing opportunities, re-mastering and more.
[1] https://entertainment.slashdot.org/story/23/11/02/177246/the-final-beatles-song-now-and-then-featuring-all-four-members-and-ai-released
[2] https://www.billboard.com/pro/the-beatles-now-and-then-first-ai-assisted-song-grammy-nomination-history/
Stem separation is a nothingburger (Score:5, Insightful)
As a musician I *deeply* dislike the sort of AI used in things like UDIO and the like. Generative slop is built on stolen art and tries to replace talent with prompts
Stem separation is not that. Its the musical equivelent of things like upscalers or whatever. Its putting nobody out of work, or acting as a substitute for talent. I have no idea how its trained, but whatever comes out is based on the musician using its own work so theres no art theft in the final result.
This isn't a problem except to people who are uncritical about clickbait.
Re: (Score:2)
> Generative slop is built on stolen art and tries to replace talent with prompts
If pop music wasn't already incredibly derivative and formulaic AI song generators wouldn't work at all. Really it's just that they pulled a Prometheus with these tools. Now anyone who fancies themselves a songwriter can actually hear their creation without having to pay someone to "compose" (which these days means using a loop based tool - seriously, watch how Taylor Swift has tracks produced for her) a track for it, and if you're not a particularly skilled vocalist, then paying someone else to sing it.
Re: (Score:2)
I dont enjoy modern pop at all. But I can assure you if it looks easy, its only cos the people who do it are good at it.
Theres a lot of training and well practiced intuition that goes into sample based music. Its not as easy as you seem to think it is.
Re: (Score:2)
> Theres a lot of training and well practiced intuition that goes into sample based music. Its not as easy as you seem to think it is.
It's not easy making money . Anyone with a rudimentary understanding of music theory and a decent size sample library can throw shit at the wall until enough of it sticks. Again though, I'm talking about producing something that's good enough for modern pop music, which you've just admitted you have little interest in.
It's a bit like saying ChatGPT isn't useful because it can't write anywhere near on the level of Shakespeare, but that's ignoring the fact that most writing isn't on Shakespeare's level to be
No double standard there (Score:1)
Regular person makes AI assisted song: Ineligible for copyright protection, no record label will sell it, and the internet will collectively groan "Oh no, not another person who thinks they're an artist because they wrote a prompt."
Famous musicians make AI assisted song: So brave!
See, the reason you or I can't use this stuff to make music anyone else has a chance of actually hearing isn't because it's not good enough, it's because we don't have the preexisting wealth, fame, and music industry connections.
Re: (Score:3)
Good art conveys intelligence in some way. See [1]the post above [slashdot.org] from a musician.
TL/DR: using AI to separate and clean up a performance of a human is just fine, because the human is still a contributor. But using generative AI to create something with trivial effort is sus.
Now, If a famous (or not famous) musician used generative AI to create something, I'd need to judge it based on how much of their own talent the project required. I'm not impressed with a work that relies completely on AI just for mimicry.
[1] https://entertainment.slashdot.org/comments.pl?sid=23517577&cid=64938355
Re: (Score:2)
> Good art conveys intelligence in some way.
I've personally witnessed Taylor Swift pack a stadium so people could hear her sing "Look what you made me do". You know, that song where the entire chorus is just, you guessed it, "Look what you made me do", ad nauseum. If ChatGPT spit out something like that, I'd ask it to try again.
Re: (Score:2)
I stand by my statement, Swifties notwithstanding. ;-P
Re: No double standard there (Score:2)
Itâ(TM)s not really the same AI, in fact itâ(TM)s just more and more computers being smart.
Generative AI is its own thing, it isnâ(TM)t the same as AI capable of isolating sounds that humans canâ(TM)t.
Re: (Score:3)
AI didn’t create anything in this instance. They used it to isolate and clean up the vocal tracks from a poor quality cassette. I’m happy they were able to pull it off and give us one final Beatles song. The George Harrison guitar tracks were recorded in the 90s but the technology wasn’t available at the time to isolate Lennon’s vocals.
Not AI Assist (Score:4, Informative)
They use software to pull out vocal track from demo and do remix later.