Was there ever a concern that Miku would make singers lose their jobs?
I don't think there was....
If not, why? What does Miku (and Vocaloid technology in general) do that makes it not-a-threat and could AI become not-a-threat in a similar manner?
=> More informations about this toot | More toots from wolf480pl@mstdn.io
@wolf480pl@mstdn.io I think the difference is that Vocaloid just still relies on the creativity of a human. But generative AI largely replaces that.
=> More informations about this toot | More toots from sigmasternchen@comfy.social
@sigmasternchen hmm... but humans do what vocaloids do too, right?
Sing the lyrics and notes written by someone else.
Does that mean this type of singing is not a creative process, merely a craft?
Is it ok to replace craftsmen but not artists?
=> More informations about this toot | More toots from wolf480pl@mstdn.io
@wolf480pl@mstdn.io It's both, I'd say. It's called "interpreter" for a reason - usually there is some transformative action involved. That's why an orchestra almost always sounds better than samples or orchestral synthesizers. I think I similar thing is going on with vocal synths.
=> More informations about this toot | More toots from sigmasternchen@comfy.social
@sigmasternchen @wolf480pl if you want a good sounding AI it requires a lot of work on the human's part, just like vocaloids. it's just that AI has a lower barrier of entry.
vocaloid is basically its own genre of music. i see AI being its own genre as well.
as a musician myself, it's hard to be mad at something that led to this.
=> More informations about this toot | More toots from beardalaxy@gameliberty.club
@beardalaxy@gameliberty.club @wolf480pl@mstdn.io Totally agree. I think using AI like synthesizers is totally fine.
The interesting part is: What about stuff like Suno that almost completely removes the creativity from the music making process. Personally I don't know. And the fact I very often just can't tell if a song was AI-generated or not is kinda disturbing. ^^"
=> More informations about this toot | More toots from sigmasternchen@comfy.social
@sigmasternchen @beardalaxy
hmm so part of the problem is that some of those AI tools can have an unlimited number of voices, while Vocaloid is just a few recognizable ones?
=> More informations about this toot | More toots from wolf480pl@mstdn.io
@wolf480pl@mstdn.io @beardalaxy@gameliberty.club I don't necessarily think so. Like: A standard synthesizer like Serum also has an unlimited number of voices.
From my perspective it's mainly a question of where the creativity comes from. (And of course the classics, like where the training data comes from, etc.)
=> More informations about this toot | More toots from sigmasternchen@comfy.social
@wolf480pl @sigmasternchen anyone can make a "vocaloid" just like anyone can train an AI on a data set. the existing vocaloids and other voicebanks have their own culture surrounding them and that can affect the decision to use a specific one.
=> More informations about this toot | More toots from beardalaxy@gameliberty.club
@wolf480pl@mstdn.io
Vocaloid never set out to deprecate physical singers altogether, neither in marketing (which is a sin of many AI companies) nor in functionality. It provided an alternative with its own pros and cons, which was particularly useful to independent music creators who could not afford to collaborate with a human vocalist, or wanted the distinct style of voice synthesis offered. It'd be like seeing Taylor Swift as a threat to all other pop musicians.
(Vocaloid also lacks the copyright/voice ownership concerns; the voice samples of Saki Fujita were used consensually.)
=> More informations about this toot | More toots from asie@mk.asie.pl
@wolf480pl@mstdn.io
A lot of LLM-based technology was already adopted without most people batting an eye; neural network-based upscaling or casual machine translation comes to mind. Heck, even generative AI was seen as a fun casual toy for the first year or two of its online presence. It is only when that technology was used to justify not making people's jobs easier, but removing entire classes of jobs (replacing them with a worse, but cheaper, end result!) did the outrage begin, really.
=> More informations about this toot | More toots from asie@mk.asie.pl
@wolf480pl@mstdn.io
In short, nobody would be taken seriously if they said "Sony can now fire all its vocalists, because they can just use Hatsune Miku". Replace "vocalists" with "artists" and "Hatsune Miku" with "Midjourney", and suddenly instead of scorn you get a VC funding round.
=> More informations about this toot | More toots from asie@mk.asie.pl
@asie good analysis.
But is "replace artists Midjourney" a credible threat because of inherent properties of Midjourney (eg. vs Vocaloid's limited selection of voices, and the training data being ethically sourced)
or is it just a human factor of being able to convince VCs, and Vocaloid with VC funding would be just as dangerous?
=> More informations about this toot | More toots from wolf480pl@mstdn.io
@asie
also
neural network-based upscaling
Yeah, waifu2x was fun.
Even though at that time I was already contemptful of AI in general.
=> More informations about this toot | More toots from wolf480pl@mstdn.io
@asie also, what if part of the problem with AI is that it pulls off a Sybil Attack?
There is one Taylor Swift, and people can generally tell a song by Taylor Swift from a song not by Taylor Swift.
There is a few Vocaloid voices and they're easy to recognize.
But a generative AI that can create millions of personas overloads our ability to attribute work to an artist, or information to a source...
=> More informations about this toot | More toots from wolf480pl@mstdn.io
@wolf480pl@mstdn.io that's actually an interesting insight
=> More informations about this toot | More toots from asie@mk.asie.pl
@wolf480pl@mstdn.io but also, another thing:
observe that popular forms of piracy and generative AI training broadly rely on the same networks of accumulating and distributing information, requiring the same type and scale of copyright infringement to operate, but people support one and oppose the other
piracy (and, say, Hatsune Miku) empower the little guy: they allow an independent artist or researcher to access information without paying a toll to corporations (say, a music label, or an academic publisher)
generative AI (and, say, cryptocurrency) typically empowers the corporations: the cost of operating them is exorbitant and only accessible to the already rich, meaning that instead of the little guy benefitting from the works funded by corporations, it's corporations benefitting from the works created by the little guy
a kind of Robin Hood mentality, essentially
=> More informations about this toot | More toots from asie@mk.asie.pl This content has been proxied by September (3851b).Proxy Information
text/gemini