333
Spotify Deleted 75 Million AI-Generated Tracks – and It’s Not Done Yet (www.gadgetreview.com)
submitted 2 months ago by to c/technology@lemmy.world
44 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[+] -19 points 2 months ago (27 children)

I wonder if this applies to Hatsune Miku or not and if not then why. Most people aren't up in arms about Hatsune Miku but is that just because they created her before there was a lot of rage about computer generated voices and personas?

  • permalink
  • fedilink
  • source
  • hideshow 27 child comments
    •
  • [–] 22 points 2 months ago (6 children)

    Hatsune Miku isn't AI generated.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [+] -9 points 2 months ago (5 children)

    Never claimed that voice was produced using NN or other algorithms typically associated with AI.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 8 points 2 months ago (3 children)

    Then why would they remove it? If you're aware that there's no connection with Hatsune Miku and AI, why are you asking as if there is?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [+] -6 points 2 months ago (2 children)

    I also never said there's no connection. Algorithms aren't just a famous democratic politician winding and grinding to a song.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 4 points 2 months ago*

    You're being disingenuous.

    If you read even the summary of the article, it makes clear that even novel AI-generated content is being allowed. So, surely, they wouldn't have an issue with what you're referring to.

    So I'm not sure why you're replying to everybody trying to maintain your stance. It just seems disingenuous.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 16 points 2 months ago (5 children)

    Synth voices are different scope to Generating actual Songs I think

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [+] -10 points 2 months ago (4 children)

    It is different as in there isn't a model or a neural network but both take real audio and algorithmically produce synthetic audio. It just seems interesting to determine one algorithm is good and another algorithm is bad.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 15 points 2 months ago (9 children)

    The ai Bros making "songs" with a Suno subscription wouldn't even be able to install vocaloid, let alone use it properly

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 9 child comments
    •
  • [+] -9 points 2 months ago (8 children)

    I'm noticing this seems to be the heart of the issue, that is the level of human involvement. If that indeed is the issue then it would be interesting to explore where the line is just from an artistic endeavour. How much human interaction or effort is enough. If for instance someone recorded all their own training data, designed their own NN, trained it, and prompted it would that be enough or not? I also get the sense that part of the issue is the underlying algorithm, that is anything involving NN is bad regardless of the circumstance.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 8 child comments
    •
  • [–] 7 points 2 months ago (6 children)

    If for instance someone recorded all their own training data, designed their own NN, trained it, and prompted it would that be enough or not?

    No, not for me. If the person involved doesn't understand how or why the sounds created are created/assembled, then they are not the artist IMO. The artist is the one with the vision, with a clear idea of how it should sound. If they are not in control of that, they are not the artist, IMO.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] -2 points 2 months ago* (last edited 2 months ago) (5 children)

    So the difference is just the understanding then? So if for instance it's a small NN and the individual does understand the input output relationship then it's art. To what level do you prefer people who use computers and electronics to understand the algorithms and electronics they are using? Like one could have a very fundamental understanding of the tools they are using to generate sounds or essentially no understanding whatsoever beyond button click makes computer / circuit go brrrrrrrrr. Like for instance I could design my own synthesizer and solder it or program a synth from scratch but I'm not sure all people who use electronics to generate sound have that level of understanding of the input output relationship. Like us someone who uses an electric piano but has absolutely no clue how the instrument operates in producing sound still an artist?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 3 points 2 months ago (4 children)

    No, this is taking it too far (and I think you might know that?).

    What I'm talking about is control, and a clear vision of what the song should sound like in advance, before recording or performing it.

    Of course, improvisation is another thing, where you lose the "in advance" aspect. But you are still in complete control in that case, of the outcome, as opposed to the case with generative AI. You are in control of every note, every sound created.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] -1 points 2 months ago* (last edited 2 months ago) (3 children)

    I am being pedantic although for sure some artists truly do have a very deep understanding of computers and electronic hardware for music production. The line is what I'm exploring. I think for instance with an NN one can most definitely have a vision, give a beatbox or musical rendition and continually edit it to their ultimate desire of what they heard in their mind. I don't see the lack of creativity in that use case although putting "make a rock song" in the prompt and hitting GO and that's it is definitely uninspired. That said that's been a debate for a long time already with electronic music and genres such as mashups or styles of music which are essentially built from nothing but samples. For decades already there's been artists mired in rumors they can't sing, play any instrument or write music aka they're just a face lip syncing or getting fully autotuned live.

    That's all to say that I'm not sure what you're saying can be applied across the board without any concern for particular use cases. I understand what you're getting at and it depends if the circumstance on hand is that or isn't, NN's or not.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 2 points 2 months ago (2 children)

    It's not about understanding how the sound is physically, or electrically generated. It's not that far down. It's about controlling the music, every note, every instrument, etc. About directly controlling the outcome.

    Electronic drums are often contentious, e.g. But if there's a person in control of every percussion stroke, it's just an instrument to me. All good.

    Even if the drums are completely programmed, there's still a person behind it who has a vision and a desire to have it sound exactly like that.

    It's not randomly generated where the "artist" presses a button and really has no idea exactly what the outcome might be.

    With regards to your idea of generating some sort of scaffolding for a song and then continually editing it, I have to agree to disagree on that point. I have a problem with the origin of the starting point.

    Even if you make a very close cover of a song that you didn't write yourself, it's still a work that had a basis in human inspiration and composition. Generative AI is not a human, even though the training data is likely of human origin.

    Ultimately it's the value we place in the amount of human input versus non-human input, in the case of works of art like music and imagery. I am not interested in non-human works. The ratio between the human and non-human input needs to be high enough for me to be interested.

    Just like in sports. If someone sets a record but was high on enhancement drugs, it doesn't count, or doesn't make it as impressive. To me at least.

    Hopefully my standpoint is made clear here.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 0 points 2 months ago* (1 child)

    So a musician who doesn't fully and directly control all the sounds in their music is not an artist? Many big artists would fall short of your definition in that someone who isn't them is controlling the lyrics, music production, final mix and so on. The artists on the label would be a massive farce as such and in actuality the producer is maybe the truest version of an actual producer.

    I still don't see the difference if someone has an idea in their mind and they are able to create that idea in the world with tool A vs tool B where tool A is a team of individuals, sound engineers, musicians, programmers and so on and tool B is a machine that can equally perform those tasks to create someone's vision. The upside is that far more people have the capacity to exercise their creative will than before.

    I do want to take time to state I am aware of the copyright issues in training some models and it is a very valid argument against certain companies and the use of certain models.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 2 months ago

    I'm having a really hard time determining if you are debating me in good faith here. It is virtually indistinguishable from willful misunderstanding.

    Anyway, I'm all for the benefit of the doubt. ❤️

    So a musician who doesn't fully and directly control all the sounds in their music is not an artist?

    Exactly. A pianist who can't control the outcome of which keys they play on their piano is not a "pianist", in my opinion. Same for any instrument, that I can think of.

    Likewise, a singer who can't control the outcome of which notes they sing, or which words they sing, is not a singer. Barring oddities like singers with Tourette's Syndrome, or a completely tone deaf (or regular deaf) person. Those types of things don't count, I suppose.

    Likewise for electronic music created with the help of a music-making program where you just drag and drop sounds. You are still in complete control of the outcome.

    Many big artists would fall short of your definition in that someone who isn't them is controlling the lyrics, music production, final mix and so on.

    Yeah, no, it's not on that level. You are both going too deep and too high with this. Again, too extreme in interpretation.

    I still don't see the difference if someone has an idea in their mind and they are able to create that idea in the world with tool A vs tool B

    Well, it's about specifics in the music, then, I guess. If your "idea" is "I want to make a rock song", you are not a creative artist. You are a prompt kiddie. That's my take. Take it or leave it.

    But if you have a musical idea, like a melody in mind, how do you prompt that? Can you play or hum a melody as a prompt? Why would you even? Are you trying to get it to finish the song for you? Why? You already have the inspiration in the form of a melody. Finish it! Know what I mean?

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 2 months ago* (last edited 2 months ago)

    Sorry for the long comment but I think it’s worth it.

    In the music industry different roles exist that change over time like they do in all industries of course but currently, there are songwriters, session players, audio engineers, producers, and artists just to mention a few. The trend is that producers need to do songwriting, audio engineering, and the actual music production for an artist nowadays. Many people want those roles to merge further with AI producing the whole thing and monetize it. That doesn’t work and I’ll explain why based on talks with people in the industry.

    If someone uses Suno, Udio, or the likes for their lyrics and/or give a chord progression along with the lyrics, they wrote and own that bit but there are tons of nuances in the composition as to which drum beat comes when, what the synth is playing (arpeggios, pedals, etc) but all that is typically what AI buddies have not even remotely thought about because “the AI will do it”.

    If they shove this into Suno, Udio, or the likes, they did not play an instrument themselves, which takes many years for it to sound good, so they did not perform the song, they did not compose the song, they did not produce the song, they did not own the full vision of the song either because they simply didn’t do it. An LLM most likely trained on stolen music is gonna do all that for them, so they don’t need to “bother” with that anymore and “just create”. It takes years and decades to master just one of those roles in the industry. Can it be rewarding to do something you couldn’t do before with AI? Hell yeah. But did you actually do anything for it? Eh, no.

    Exactly for that reason you don’t actually own the piece. In fact, by law you don’t. That’s why you cannot submit it for sync licensing. Agents won’t take it. They don’t want to get sued because AI music made like that cannot be owned. It’s ripped from a bunch of other artists who might potentially sue them.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 13 points 2 months ago (3 children)

    I think the Vocaloids are more similar to weird instruments that are still played by humans, than to generative AI song synthesis.

    https://upload.wikimedia.org/wikipedia/de/e/eb/Vocaloid1_-_Screenshot.jpg

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] -1 points 2 months ago (2 children)

    I think you're saying. It's the level of human involvement that determines if it's good or bad? If so it would be interesting to explore exactly where the cutoff is.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 8 points 2 months ago (1 child)

    All I'm saying here is that Vocaloid is essentially an electronic instrument, like a keyboard, or a theremin, or a vocoder, or a synthesizer. Vocaloid is not very similar to AI generation, so there is no particular reason to assume that Spotify would them together.

    I'm not trying to comment on what I find good or bad, nor to get into the fundamental debate on what music or art is.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] -5 points 2 months ago

    Yeah you haven't made any statements but I think we are both aware of the sentiments or feelings of people on the subject. Personally wherever there are strong feelings or clashing perspectives there's a source of exploration and artistic creation.

    Like for instance if someone created their own sample database, designed their own NN, trained it, and prompted it would that be enough human involvement or maybe if they had it generate then samples to be fed into a vocaloid then it would be enough or maybe for some that's still not valid music.

  • permalink
  • fedilink
  • source
  • parent