AI & the future of Music

[Goredho 17412]

I thought about putting this in the "Share Your Music Thread" because I have been using some AI stuff on some music there recently, but I think this is probably more interesting to a broader audience including people who just consume music.  Gonna tag people active in that thread @sidis, @jimmyjazz, @G650.

https://mixmag.net/read/grimes-software-elftech-mimic-voice-percent-royalties-news

Quote

Grimes is encouraging fans and producers to “take advantage” of her vocals by launching a new software that mimics her voice.

Elf.Tech, the AI voice software currently in beta mode, allows users to record and upload their own voice which will then spit it out in the sound of Grimes’ when run through the software.

Alongside the launch of Elf.Tech, Grimes announced that she would offer a 50% royalty split to anyone who collects them on a track using her AI-generated voice.

https://mixmag.net/read/kito-and-grimes-share-new-ai-collaborative-track-made-using-grimes-ai-software-elf-tech-news/

Quote

Australian-born LA-based Dj and producer Kito has used Grimes’ Elf.Tech software, mimicking her voice, in a new demo ‘Cold Touch' — becoming Grimes’ first, official AI release.

The track was teased across Kito’s socials last week, and caught praise from Grimes herself — who called it a “masterpiece” and garnered an overwhelmingly positive response from fans and fellow artists.

There's a lot here that's interesting.  Grimes using AI to essentially scale out her ability to perform and earn royalties.  I think her music sucks, but that's kinda brilliant.  The possibility to resurrect dead performers ("anyone want a new Prince track?").  Potential threat to session artists (what if the vocals on Pink Floyd's "The Great Gig in the Sky" could have been generated by AI vs the impromptu performance of a human?).  Could we ever reach a point where AI might make exercising human creativity... pointless?

The song in question.  This is not my cup of tea, and the vocals certainly don't sound natural, but this is where we're going.  Generative AI's arrival is going to make the future of music interesting.

 

[G650 14267]

58 minutes ago, Goredho said:

but I think this is probably more interesting to a broader audience including people who just consume music.

 

Yeah, I think you right.

 

My take initially (as in generally not specifically your post) is there will be people affected by this in myriad ways. How, it's hard to say yet. For me personally, I just can't see an impact. I hate to say I just don't care, but, .....I kinda just don't care. That's not to say I won't be listening to something in the future AI generated, but rather it isn't going to give me the same response. Quite frankly I find a ton of current music might as well have been AI. It's so homogeneous in it's production aesthetic it doesn't really trigger much in me. Not all for sure, there's plenty of good stuff, I am not one of those yell at clouds types ( at least in this regard). But in the main, it's all bland sameness. So bring on our new AI overlords.

[Helobious 9463]

It’s hard to tell. For example I think the song OP posted sounds like shit, because I just don’t like that kind of music. I thought the whole point of AI was to do stuff better than a human ever could. But for subjective things like art & music, that becomes irrelevant because it’s all based on the taste of the consume anyway. You could make a perfectly AI-generated Incubus song, I’m still not going to listen to it because I don’t like Incubus.

[G650 14267]

I think the previous post was AI given the poster's history

[Goredho 17412]

12 minutes ago, G650 said:

 

Yeah, I think you right.

 

My take initially (as in generally not specifically your post) is there will be people affected by this in myriad ways. How, it's hard to say yet. For me personally, I just can't see an impact. I hate to say I just don't care, but, .....I kinda just don't care. That's not to say I won't be listening to something in the future AI generated, but rather it isn't going to give me the same response. Quite frankly I find a ton of current music might as well have been AI. It's so homogeneous in it's production aesthetic it doesn't really trigger much in me. Not all for sure, there's plenty of good stuff, I am not one of those yell at clouds types ( at least in this regard). But in the main, it's all bland sameness. So bring on our new AI overlords.

Agree that the state of the AI art now isn't that evocative/emotionally moving.  Like that Kiki + GrimesAI song above... I don't get it.  I mean, I get it from Grimes' perspective as a way to make more money, and from Kiko's perspective as a way to attract more attention, but as a listener, the music isn't remarkable to me in any way.  But that may be because I'm one of those yell at clouds types

Which is a possible outcome.  If this is it, and all AI ever does is produce bland, least-common denominator trash music derivative of what has already been, then human creatives won't be replaced.  AI will generate crappy throw away music the equivalent of toys at the Dollar General.  It may put people writing jingles for advertising out of a job, but you still need humans pushing music's evolution.  Using a historical example, AI in its present form couldn't do something like take blues, R&B, doowop and country swing as input and produce rock & roll as an output.

Maybe that's a landmark for being scared.  When/if an AI can ever do something uniquely creative vs solely derivative.

[G650 14267]

I have to admit, I had no idea who Grimes was until Elon Musk bought Twitter.

[jimmyjazz 45989]

Wait . . . am I misunderstanding something here?  The "sound" of one's voice is not (to my knowledge) protected by IP law.  If I create a new song that sounds exactly like Sting wrote and recorded it, BUT HE DIDN'T, why should he share royalties?

[Goredho 17412]

3 minutes ago, jimmyjazz said:

Wait . . . am I misunderstanding something here?  The "sound" of one's voice is not (to my knowledge) protected by IP law.  If I create a new song that sounds exactly like Sting wrote and recorded it, BUT HE DIDN'T, why should he share royalties?

I have no clue, but I expect there is no definitive answer.  What you describe is just now coming into the realm of possibility.  I expect it will get hashed out in the courts soon enough.

That may be part of Grimes’ move here.  Preemptively making this possible with her voice using her AI may allow her some measure of control over access, licensing and terms of use.

But if I go make my own AI seeded with just grimes vocals to make my own virtual Grimes for the studio?  I don’t know what that means from a legal standpoint.

[Deej 76842]

1 hour ago, Goredho said:

But if I go make my own AI seeded with just grimes vocals to make my own virtual Grimes for the studio?  I don’t know what that means from a legal standpoint.

Fuck around and find out. 

[tbone_ 14132]

I don’t like it. I’m way more of a low tech, just let humans make music type.

But hell I guess even low tech was new tech at some point. Someone probably thought the invention of the lute was going to ruin music forever. So maybe I am yelling at clouds.

[G650 14267]

1 minute ago, tbone_ said:

Someone probably thought the invention of the lute was going to ruin music forever

But were they wrong

[Goredho 17412]

43 minutes ago, tbone_ said:

I don’t like it. I’m way more of a low tech, just let humans make music type.

But hell I guess even low tech was new tech at some point. Someone probably thought the invention of the lute was going to ruin music forever. So maybe I am yelling at clouds.

I think there is a big difference, though.  A lute still requires human skill and creativity to produce music.  We’re moving towards a place where you will be able to give an AI a set of lyrics, a song track, information about the song’s structure like the key and chord chart, and some guidance for the voice and style of performance and the AI spits out a vocal track for you, you iterate with the AI to a final performance you like, and then you download it and add it to your mix.

Assuming a high enough degree of quality is eventually achieved (it may never), we will have “progressed” to not needing the human skill or the human creativity for a compelling vocal performance.

Thats fundamentally different than the invention of the lute, which still required a human to operate and a human mind to use creatively.  We may never get to that point, as @G650 pointed out, it may just forever be derivative and bland.  But there are products on the market today for what I described.  They are lacking, but there is going to at least be the attempt to provide producers with the ability to not need  human musical skill or creativity to make music.  If the attempt succeeds, creating and delivering a song will be:

Generate a slow blues tune, say at a tempo of 80 bpm in 4/4 time.  It should have guitar bass and keyboard in the mix.  It should be in the key Am.  It should have an intro, verse, chorus, verse, chorus, bridge, chorus x2 structure.

Great, but I don’t like the chord used for the 2 bar of the chorus, can you try something minor there?

Can you make the guitar a little brighter, maybe use a telecaster?

Can you generate a set of lyrics for this song about the heartache of outliving the love of your life?

<iterate on the lyrics with the AI until happy>

Ok, apply these lyrics to the song as a vocal performance.  Use a husky female voice, perhaps like Amy Winehouse.

<iterate on the vocal performance with the AI until happy>

<iterate on the mix with the AI until happy>

<iterate on the master with the AI until happy>

AI write a social media post about this fantastic new song I just composed, arranged, performed, recorded, mixed and mastered.  Include a description of the song.  Push it to all my social media accounts.

[tbone_ 14132]

[jimmyjazz 45989]

@Goredho but what you just described is a scenario in which some human had a clear picture of what they wanted, and worked to get it.  I'm not saying it's a good thing, it's not, but it's not random, either.

Of course, the reality could be a lot worse than what you described.

[Goredho 17412]

34 minutes ago, jimmyjazz said:

@Goredho but what you just described is a scenario in which some human had a clear picture of what they wanted, and worked to get it.  I'm not saying it's a good thing, it's not, but it's not random, either.

Of course, the reality could be a lot worse than what you described.

Yep, and one could have the view that ultimately it’s just the music that matters, and instruments (including the human voice) and the ability to utilize them to realize the “vision” or idea of a song is actually an impediment to creativity.  An outdated concept that was necessary to manufacture music before AI.  That removing the barrier of tools and skills necessary to use them between the music someone hears in their head and the music that comes out of a speaker will empower more creativity.

Like, I can’t sing.  Or play the fiddle.  Or the drums.  Or be a great audio engineer and professionally master my tracks.  And as a semi-pro hobbyist, I don’t have the budget to afford humans who are.  For someone like me, AI is going to help fill in those gaps and allow me to produce music I would never be able to otherwise.  I’ll be exercising my personal skill & creativity with guitars, bass and keyboards, and delegating much of the rest of what I hear in the song in my head to AI.

That’s going to be pretty cool.

Edited May 20, 2023 by Goredho

[G650 14267]

5 minutes ago, Goredho said:

nstruments (including the human voice) and the ability to utilize them to realize the “vision” or idea of a song is actually an impediment to creativity

Eh, I'm not sure I buy that. The creativity is inside you. However you realize it is certainly  a discussion but it exists independently of any external factors.

[Goredho 17412]

22 minutes ago, G650 said:

Eh, I'm not sure I buy that. The creativity is inside you. However you realize it is certainly  a discussion but it exists independently of any external factors.

Right, so if something enables you to express your internal creativity with no need for the 10,000 hours to build mechanics and technique to manipulate a tool to realize your vision, one could take the view that AI will be unleashing human creativity, not ending it.

If we get to the point where my described scenario is real, the 10,000 hours someone used to spend honing their technical skills with an instrument might be spent listening to music and developing the catalog of sounds in their brain so they can guide AI creatively.

[G650 14267]

10 minutes ago, Goredho said:

Right, so if something enables you to express your internal creativity with no need for the 10,000 hours to build mechanics and technique to manipulate a tool to realize your vision, one could take the view that AI will be unleashing human creativity, not ending it.

If we get to the point where my described scenario is real, the 10,000 hours someone used to spend honing their technical skills with an instrument might be spent listening to music and developing the catalog of sounds in their brain so they can guide AI creatively.

But it's not though. It's just giving you a backing track.

[Goredho 17412]

1 hour ago, G650 said:

But it's not though. It's just giving you a backing track.

I'm playing Devil's advocate with the past couple posts.  I'm not sure I fully buy the argument, either, but I think there is some logic to it.  Like, where does the creativity lie.  Is it in the creative thought?  The act that makes the creative thought "real"?  Some combination of the two?  If one takes the view that the creativity lies solely with the creative idea/conception, then all an AI is doing is rendering your thought so that others can hear it, too.

Like, if Beethoven had access to a future, further-advanced AI, and he described the sounds he heard in his head to it ("it should go BUM-BUM-BUM-BUMMMMM really dramatically!"), and after a process of iteration and refinement with the AI, it spit out the sheet music for all the instruments as well as a sonic rendering of a full orchestra playing his 5th symphony that was indistinguishable from a performance by the Royal Philharmonic Orchestra... was Beethoven's creativity any less exercised than how he actually composed his 5th symphony?   Working through the music within his head and at a piano, iterating on it, refining it, then writing the musical notation for all the parts and finally having a full orchestra play it?

Edited May 20, 2023 by Goredho

[G650 14267]

7 minutes ago, Goredho said:

I'm playing Devil's advocate with the past couple posts.  I'm not sure I fully buy the argument, either, but I think there is some logic to it.  Like, where does the creativity lie.  Is it in the creative thought?  The act that makes the creative thought "real"?  Some combination of the two?  If one takes the view that the creativity lies solely with the creative idea/conception, then all an AI is doing is rendering your thought so that others can hear it, too.

Like, if Beethoven had access to a future, further-advanced AI, and he described the sounds he heard in his head to it ("it should go BUM-BUM-BUM-BUMMMMM really dramatically!"), and after a process of iteration and refinement with the AI, it spit out the sheet music for all the instruments as well as a sonic rendering of a full orchestra playing his 5th symphony that was indistinguishable from a performance by the Royal Philharmonic Orchestra... was Beethoven's creativity any less exercised than how he actually composed his 5th symphony?   Working through the music within his head and at a piano, iterating on it, refining it, then writing the musical notation for all the parts and finally having a full orchestra play it?

Yeah I get you.

 

The Beethoven example though is kind of what I mean, I don't think it would produce it exactly as he had it in his head, it would produce a Beethoven's 5th themed piece. Kind of like how Sunset Blvd doesn't have actual Irish or Country bars, but Irish and Country themed bars.

[jimmyjazz 45989]

I doubt (had he been able to hear) Beethoven would have thought any particular orchestra playing the 5th would have been perfectly capturing his vision, either.

Honestly, I'm not sure I'm creative enough to really understand what it means to try to pull great ideas out of the ether and fix them in time and space.  I mean, sure, I'm a dumbass who writes songs with a bit of wit and craft but I'm not exactly setting the world on fire.  I don't know what it means to live in the head of Aaron Copland or Miles Davis or Beethoven or whoever.  It's just not something I'm adept enough at to really "get".

[Goredho 17412]

Well, me either, but I imagine things on instruments I can’t play.  I’m working on a folksy/Americana type piece.  It’s all on guitar at the moment, but I hear a fiddle counterpoint melody over it.  The notes, things like slides and swells.  I can play the notes I hear in my head on guitar, but I can’t play the fiddle or carry out its nuances on something like a guitar.

If I describe this to an AI to render a track, the notes, the attacks, the swells, the slides and the vibratos…. and it carried that out to generate a track that I incorporate into the song, is it that different than giving similar instructions to a session player?  Am I being any more or less creative if I were to use an AI vs a session player in this way?

Edited May 21, 2023 by Goredho

[jimmyjazz 45989]

8 minutes ago, Goredho said:

Well, me either, but I imagine things on instruments I can’t play.  I’m working on a folksy/Americana type piece.  It’s all on guitar at the moment, but I hear a fiddle counterpoint melody over it.  The notes, things like slides and swells.  I can play the notes I hear in my head on guitar, but I can’t play the fiddle or carry out its nuances on something like a guitar.

If I describe this to an AI to render a track, the notes, the attacks, the swells, the slides and the vibratos…. and it carried that out to generate a track that I incorporate into the song, is it that different than giving similar instructions to a session player?  Am I being any more or less creative if I were to use an AI vs a session player in this way?

Not really, other than the slight discrepancies between a human being and an AI version of a human being.

Sometimes the beauty is in the flaws, and (so far) I have never seen a credible method to create "human" flaws (randomness in drum machine tracks, etc.).   But, AI will probably surprise.

 

[Goredho 17412]

7 minutes ago, jimmyjazz said:

Not really, other than the slight discrepancies between a human being and an AI version of a human being.

Sometimes the beauty is in the flaws, and (so far) I have never seen a credible method to create "human" flaws (randomness in drum machine tracks, etc.).   But, AI will probably surprise.

 

For sure.  And all of this is a hypothetical based on a premise that AI will be able to generate a performance comparable to a human.  Totally agree with the human flaws to a performance, where not everything is perfectly aligned in a grid.

[G650 14267]

I do think you guys are underestimating a Beethoven or Mozart's ability to translate specificity to their scores, and an orchestral players ability to execute.

 

Now I think if you input a score into an AI, it could probably produce the performance very closely, but that to me is a different conversation.

[Goredho 17412]

2 hours ago, G650 said:

Now I think if you input a score into an AI, it could probably produce the performance very closely, but that to me is a different conversation.

You're right, it is kind of a different conversation, and talking through it here, I think the creativity lies with the human in that scenario.  The technical skill to render the creation is in the AI, but the actual conception is purely human.  It eliminates the need for mechanical/performative skill, but not the need for human creativity.

I took a couple fucking gummies.  You guys probably don't want to respond, or I'm highly likely to write a THC fueled dissertation on AI.  

[G650 14267]

7 hours ago, Goredho said:

I took a couple fucking gummies.  You guys probably don't want to respond, or I'm highly likely to write a THC fueled dissertation on AI.  

Now I'm intrigued

[Goredho 17412]

Too late, stone sober now.

[G650 14267]

[TwiceHorn 31509]

This is interesting from the IP standpoint, and that in some ways ties into the creativity/compositional aspect under discussion.

I think jimmy is right that there's no basis in creativity/originality (copyright) or conception of an idea/inventorship (patent) for the sound of a voice, or an instrument for that matter. 

There are trademarks in particular sounds, e.g. Harley has trademarked the sound of their particular v-twin engine, NBC has trademarked the "peacock sound."  That is dependent not on creativity, originality, or inventorship, but on the public associating the sound with a source.  Trademarks can be licensed without regard to their actual validity as trademarks.  NIL is a great example of that, meaning player license their image successfully without regard to whether the public actually associates their image with anything (of course, the fact that they can license it implies some level of recognition/association in and of itself).  People license invalid/unenforceable IP all the time.

As has been discussed, the current state of the law of copyright and patent in the US denies authorship/inventorship to non-human entities, so no AI inventors.

But the prompts to an AI machine could be the source of authorship/inventorship, legally.  So, by extension of that concept, prompts to an AI machine that generate music are entitled to composition credit, if a diminished form of it.  Programming a certain voice involves no real creativity/authorship, possibly inventorship as I imagine it is not a mean feat to program a voice and all its nuances as a response in an AI machine.

 

[G650 14267]

Jesus man.

[Goredho 17412]

39 minutes ago, TwiceHorn said:

This is interesting from the IP standpoint, and that in some ways ties into the creativity/compositional aspect under discussion.

I think jimmy is right that there's no basis in creativity/originality (copyright) or conception of an idea/inventorship (patent) for the sound of a voice, or an instrument for that matter. 

There are trademarks in particular sounds, e.g. Harley has trademarked the sound of their particular v-twin engine, NBC has trademarked the "peacock sound."  That is dependent not on creativity, originality, or inventorship, but on the public associating the sound with a source.  Trademarks can be licensed without regard to their actual validity as trademarks.  NIL is a great example of that, meaning player license their image successfully without regard to whether the public actually associates their image with anything (of course, the fact that they can license it implies some level of recognition/association in and of itself).  People license invalid/unenforceable IP all the time.

As has been discussed, the current state of the law of copyright and patent in the US denies authorship/inventorship to non-human entities, so no AI inventors.

But the prompts to an AI machine could be the source of authorship/inventorship, legally.  So, by extension of that concept, prompts to an AI machine that generate music are entitled to composition credit, if a diminished form of it.  Programming a certain voice involves no real creativity/authorship, possibly inventorship as I imagine it is not a mean feat to program a voice and all its nuances as a response in an AI machine.

 

Interesting.  I know most of the AI stuff I use grants a license for its output suitable for commercial purposes.  But it sounds like from the above that may not even be necessary?  Once a creative piece is generated from AI prompts by a human, the owner/operator has no right?

[TwiceHorn 31509]

12 minutes ago, Goredho said:

Interesting.  I know most of the AI stuff I use grants a license for its output suitable for commercial purposes.  But it sounds like from the above that may not even be necessary?  Once a creative piece is generated from AI prompts by a human, the owner/operator has no right?

Well, the current state of the law is that authorship and inventorship of the output of an AI machine will not be recognized as valid, with the author or inventor as the AI machine.

It seems the most likely author or inventor would be the "prompter" or inputter.  Although another theory holds that it would be a "group effort" of the AI machine designers and the inputters.  Under the first theory, no license needed, under the second theory, license is probably needed.

[BrickHorn 18246]

20 minutes ago, TwiceHorn said:

It seems the most likely author or inventor would be the "prompter" or inputter.

I thought the Copyright Office guidance differentiates between creations of AI based on prompts (not copyright able) versus just implementing a human’s artistic vision (copyrightable). In other words, a composition would only be copyrightable if it is the artistic work of a human being.

[TwiceHorn 31509]

1 hour ago, BrickHorn said:

I thought the Copyright Office guidance differentiates between creations of AI based on prompts (not copyright able) versus just implementing a human’s artistic vision (copyrightable). In other words, a composition would only be copyrightable if it is the artistic work of a human being.

I believe you are correct. https://copyright.gov/ai/ai_policy_guidance.pdf

I was focused more on the patent regime, where the courts have uniformly ruled in the US and elsewhere, that AI machines may not be inventors and left everything else open.  Of course, the standards there are a little different.  An invention may be otherwise patentable, with the open question of who are the proper inventors.  The copyight question goes a little deeper as to whether the material is even eligible for copyright if it lacks human authorship, regardless of who are the proper authors.

Edited May 21, 2023 by TwiceHorn

[BrickHorn 18246]

3 hours ago, TwiceHorn said:

The copyight question goes a little deeper as to whether the material is even eligible for copyright if it lacks human authorship, regardless of who are the proper authors.

Which will create all kinds of weird incentives. 

[Goredho 17412]

Took this line of our conversation and approached ChatGPT with it.  Who needs high priced legal advice, anyway?  This is long, but I think pretty interesting/fascinating.

Me
I would like to talk about how do current laws and court rulings in the United States apply to AI generated content.  Specifically musical content.  Can we do that?

ChatGPT
Certainly! I can provide you with some general information on how current laws and court rulings in the United States apply to AI-generated content, including musical content. However, please keep in mind that I am an AI language model, and my responses should not be considered as legal advice. If you require specific legal advice, it is best to consult with a qualified attorney who specializes in intellectual property and copyright law.

In the United States, copyright law protects original works of authorship, including musical compositions. Copyright law grants exclusive rights to the creators of a work, such as the right to reproduce, distribute, publicly perform, and create derivative works based on the original composition. These rights are typically granted to the human creators of the musical content.

When it comes to AI-generated musical content, the legal implications can be complex. While AI itself cannot hold copyright, the question arises as to who owns the copyright in the output generated by AI systems. It often depends on the level of human involvement in the creation process.

If a human actively contributes to the creation and direction of the AI-generated music, they may be considered the copyright owner of the resulting composition. However, if the AI system independently generates the music without substantial human involvement, questions regarding copyright ownership become more complicated.

To date, there have been no specific court rulings or legislation in the United States addressing copyright ownership of AI-generated content. However, existing copyright laws could potentially be applied to determine ownership and infringement claims.

Additionally, fair use is an important aspect of copyright law. Fair use allows limited use of copyrighted material without permission from the copyright holder. It encompasses uses such as commentary, criticism, parody, and transformative uses. The application of fair use to AI-generated musical content will depend on the specific circumstances and factors involved in each case.

It's worth mentioning that legal frameworks surrounding AI-generated content are still evolving, and there are ongoing discussions on the need for potential updates to copyright laws to address AI-generated works.

To fully understand the legal landscape and implications of AI-generated musical content, it is crucial to consult with legal experts who specialize in intellectual property law and keep track of any new developments in legislation and court rulings.

Me
So I am going to stick to music for any examples.  Would it be possible to train an AI solely on the works of another artist to produce output that sounds much like the original artist?  For example, if I were to train an AI with solely Prince vocal recordings, would it be possible for the AI to generate media that sounded very close to a Prince vocal recording?

ChatGPT
Yes, it is possible to train an AI on the works of a particular artist, such as Prince, in order to generate output that sounds similar to the original artist's style. This process is often referred to as "style transfer" or "generative modeling." By training the AI on a large dataset of Prince's vocal recordings, the AI can learn the patterns, nuances, and characteristics of Prince's singing style.

There have been advancements in the field of AI, particularly in the area of deep learning and neural networks, that have enabled the development of models capable of generating highly realistic and convincing outputs, including music and vocal recordings.

However, it's important to note that AI-generated content intended to mimic or imitate the style of a specific artist raises legal and ethical considerations. As mentioned earlier, copyright laws protect original works, and using an AI to replicate a specific artist's style could potentially infringe upon their copyright. If the AI-generated output closely resembles the original artist's work and is distributed without proper authorization or permission, it could potentially lead to legal issues.

Furthermore, there are ethical concerns surrounding the use of AI to replicate an artist's style without their consent. Artists often have personal and creative reasons for maintaining control over their work and may object to the creation of AI-generated content that imitates their style or voice.

It's important to approach the creation and use of AI-generated content ethically, respecting the rights and artistic integrity of the original creators. Consulting with legal experts and obtaining proper permissions and licenses from the relevant copyright holders is advisable when working with AI-generated content based on existing artists' works.

Me
You already jumped to my next question.  If I used an AI trained solely on Prince's vocal recordings to produce my own works that sound like Prince, but are original recordings otherwise -- completely original songs with original lyrics and melodies -- why would Prince (or his estate) be entitled to a right to the composition?  The composition is original, it would seem the copyright would be safe.  But the performance by Prince AI of the copyrighted material might be immoral, unethical or illegal.

ChatGPT
If you were to use an AI trained solely on Prince's vocal recordings to produce original songs with original lyrics and melodies, the copyright in the composition itself would indeed be separate from the performance by the Prince AI. The composition, consisting of the original lyrics and melodies, would be considered your own original work and could potentially be protected by copyright.

However, it's important to note that the performance aspect, which involves the Prince AI mimicking the vocal style of Prince, might raise legal and ethical considerations. While the composition may be original, the AI-generated performance could potentially infringe upon the rights associated with Prince's persona and voice.

Personality rights and rights of publicity are legal concepts that grant individuals control over the commercial use of their likeness, voice, or other distinctive aspects of their persona. These rights are often associated with celebrities and public figures, including their unique vocal style.

Even if the composition itself is original, using an AI to replicate the performance of a specific artist like Prince, without proper authorization or permission, could potentially infringe upon the rights associated with Prince's persona. Prince, or his estate, could argue that the AI-generated performance misappropriates his distinctive vocal style and violates his personality rights.

Additionally, there may be ethical considerations surrounding the use of AI-generated performances that replicate an artist's style without their consent. Artists often have personal and creative reasons for wanting to maintain control over their own voice and how it is used.

While the legal and ethical landscape surrounding AI-generated content is still evolving, it's important to approach the creation and use of such content with care and consideration for the rights and artistic integrity of the original creators. Consulting with legal experts is advisable to navigate the complex legal implications associated with AI-generated performances based on existing artists' works.

 

Me
If Prince's estate created the Prince AI trained solely on Prince's vocals that could take original performance descriptions and generate a vocal performance that sounded as if Prince himself had sung it, could the estate enforce terms of its use?  For instance, could they require sharing in 50% royalties and earnings of any songs that use Prince's AI vocals, and enforce that?
 

ChatGPT
The enforceability of specific terms and conditions set by Prince's estate regarding the use of Prince AI vocals would depend on several factors, including the legal agreements in place, the jurisdiction in which the agreements are enforced, and the specific circumstances surrounding the use of the AI-generated vocals.

If Prince's estate were to create and own the Prince AI trained solely on Prince's vocals, they could potentially establish licensing agreements or contracts that govern the use of the AI-generated vocals. These agreements could outline the terms and conditions for the use of the AI vocals, including the requirement to share royalties and earnings.

The enforceability of such terms would depend on various legal factors, including the nature and validity of the agreements, any relevant intellectual property laws, and the specific rights granted to the estate over Prince's voice and likeness.

In some jurisdictions, personality rights or rights of publicity may provide legal protection for an individual's voice and likeness, allowing them to control and monetize their use. If Prince's estate has valid and recognized rights over Prince's voice, they may have the ability to enforce certain conditions on the use of the AI-generated vocals.

It's important to note that legal frameworks surrounding AI-generated content are still evolving, and the enforceability of specific terms and conditions can vary. It would be advisable to consult with legal experts who specialize in intellectual property law and have a deep understanding of the specific jurisdiction and circumstances involved to determine the enforceability of such terms.

Additionally, it's worth considering the ethical implications of using AI-generated vocals that mimic a deceased artist's voice. Public perception and the potential impact on the artist's legacy should also be taken into account when making decisions about the use and licensing of such AI-generated content.

 

Me
Lets say this is in the United States.  So use the US as the jurisdiction and amend your previous answer.

ChatGPT
In the United States, the enforceability of specific terms and conditions set by Prince's estate regarding the use of Prince AI vocals would still depend on various legal factors. While I can provide some general insights, it's important to consult with legal experts who specialize in intellectual property and contract law to obtain specific advice tailored to your situation.

In the United States, intellectual property rights, including rights of publicity and personality rights, can vary from state to state. Some states have specific laws that recognize and protect an individual's voice, likeness, or persona, while others rely on common law doctrines or case precedents.

If Prince's estate has established rights of publicity or personality rights over Prince's voice, they may have the ability to enforce certain conditions on the use of the AI-generated vocals. However, the specific terms and conditions, as well as the enforceability of those terms, would depend on the agreements or contracts in place between the estate and the users of the AI vocals.

For the estate to require sharing in royalties and earnings from songs that use Prince's AI vocals, it would generally require contractual agreements between the estate and the users. The enforceability of these agreements would depend on the validity of the contracts, including considerations such as offer, acceptance, consideration, and the absence of any unconscionable or illegal terms.

It's worth noting that enforcing contractual terms can involve legal proceedings, and the outcome will ultimately depend on the specific facts and circumstances of each case, as well as the interpretation of relevant laws by the court.

Therefore, in the context of the United States, while the estate may have a basis to establish certain conditions on the use of the AI vocals if they hold valid rights, the enforceability of those conditions would still require careful consideration of the specific legal agreements, applicable state laws, and potential legal challenges that may arise.

Once again, it is highly recommended to consult with legal experts who specialize in intellectual property and contract law to obtain precise and up-to-date advice in the United States.

 

Me
Ok, amend your previous answer to use a jurisdiction of the state of Colorado, located in the United States.

ChatGPT
As an AI language model, I don't have real-time access to current state-specific laws or the ability to provide jurisdiction-specific legal advice. Laws can vary between jurisdictions, and it is crucial to consult with legal professionals who are familiar with the laws of the specific jurisdiction, such as the state of Colorado, for accurate and up-to-date information.

That being said, I can provide some general insights regarding the United States as a whole. In the United States, intellectual property rights, including rights of publicity and personality rights, can vary from state to state. Some states have specific statutes or case precedents that recognize and protect an individual's voice, likeness, or persona, while others rely on common law doctrines.

If Prince's estate were to enforce terms and conditions, such as requiring sharing in royalties and earnings from songs that use Prince's AI vocals, in the state of Colorado, it would likely depend on the state's specific laws and legal precedents.

In the absence of comprehensive state laws regarding personality rights, Colorado may rely on common law doctrines or general principles of contract law. This means that the enforceability of the terms and conditions would depend on the specific agreements or contracts between Prince's estate and the users of the AI vocals, as well as general contract law principles such as offer, acceptance, consideration, and the absence of any unconscionable or illegal terms.

To obtain accurate and reliable information on the enforceability of specific terms and conditions in the state of Colorado, it is strongly recommended to consult with legal professionals who specialize in intellectual property and contract law within the state. They will have the most up-to-date knowledge and can provide you with precise guidance tailored to the laws of Colorado.

 

Me
Ok, so lets keep all examples that might depend on jurisdiction limited to the United States.

ChatGPT
Certainly! I will focus on providing information and examples within the context of the United States as a whole, rather than specific state jurisdictions. Please note that the following examples should be considered for illustrative purposes only and are not legal advice.

Licensing AI-generated vocals: Prince's estate could choose to license the AI-generated vocals for use in original compositions. They could require users to obtain a license and pay royalties or earnings based on the terms negotiated in the licensing agreement.

Terms of use and restrictions: Prince's estate might enforce certain terms of use for the AI-generated vocals. For instance, they could require attribution, specify limitations on the extent of use, or prohibit certain types of commercial or exploitative uses.

Control over derivative works: Prince's estate could claim control over derivative works created using the AI-generated vocals. This would allow them to exercise rights over adaptations, remixes, or other creative works that incorporate the AI vocals, potentially requiring permission and sharing in royalties or earnings.

Moral rights and artistic integrity: While the United States does not have a comprehensive system for moral rights, Prince's estate could advocate for protection of the integrity and reputation of the AI-generated vocals. They might seek to enforce conditions that ensure the AI vocals are used in a manner consistent with Prince's artistic vision and legacy.

Contractual obligations: Prince's estate could establish contractual agreements with users of the AI-generated vocals to outline specific obligations, such as quality control, exclusivity, or restrictions on the duration of use.

It's important to note that the enforceability of these examples would still depend on various factors, including the specific agreements, applicable federal laws, potential fair use considerations, and other legal doctrines. Consulting with legal professionals who specialize in intellectual property and contract law in the United States would be essential to understand the intricacies and enforceability of such terms and conditions.

 

Me
Thanks for the summation.  Now, lets say I as a different legal entity than Prince's estate, took all of Prince's recorded works, and used them to train an AI myself, lets call it the Outlaw Prince AI.  If I created a work with it that included an original vocal performance specified by me, including the exact notes being sung, I assume Prince's estate who own the PrinceAI would have a claim to stop distribution of the song using an unauthorized performance, correct?  But it sounds like the copyright of the original work, the notes being sung, would be held by me as the creative author.  Would you say that seems correct?

ChatGPT
If you, as a different legal entity, trained an AI using Prince's recorded works and created the Outlaw Prince AI, the legal implications surrounding the use of the AI-generated content would be complex. While I can provide some general insights, please note that copyright law can be intricate, and it is advisable to consult with legal professionals who specialize in intellectual property law for specific advice tailored to your situation.

In this scenario, if you use the Outlaw Prince AI to create a work that includes an original vocal performance specified by you, it is important to consider two aspects: the underlying composition and the performance itself.

Underlying composition: The composition, which includes the original notes being sung and other musical elements, could potentially be considered an original work authored by you. As the creator of the composition, you may hold the copyright to the original work, assuming it meets the threshold of originality and is not infringing on any preexisting works.

Performance using the AI-generated vocals: The performance aspect involves the use of the AI-generated vocals that were trained using Prince's recorded works. Prince's estate, who owns the Prince AI, could potentially claim that the unauthorized use of the AI-generated vocals infringes on their rights to control the commercial use of Prince's voice and likeness. They may argue that the unauthorized use misappropriates Prince's distinctive vocal style and violates his personality rights.

In such a case, Prince's estate may have the right to take legal action to stop the distribution of the song using the unauthorized performance, based on claims of infringement of personality rights or rights of publicity associated with Prince's voice.

It's worth noting that the specific facts, circumstances, applicable laws, and any relevant agreements or licenses could influence the outcome. Intellectual property laws, including copyright and personality rights, can vary, and the interpretation of these laws can be subject to legal proceedings and judicial discretion.

To obtain accurate and reliable legal advice, it is strongly recommended to consult with legal professionals who specialize in intellectual property law, particularly in the jurisdiction where the potential legal disputes may arise. They will have the expertise to provide specific guidance based on the applicable laws and the particular circumstances of your case.

 

Me
So using this example scenario, while it may be unethical or immoral to do this, I might use the following tactic to generate attention for my original copyrighted work.  Train my own unlicensed Outlaw Prince AI with all of Prince's vocals.  Get it to provide a vocal performance for my original work that I hold the copyright for.  Distribute that and advertise it as a song featuring the "Second Coming of Musical Jesus -- Prince!" to generate attention.  When Prince's estate eventually contacts me to stop distribution of the song, I immediately comply, and release a new version of the same composition that I hold the copyright for but with a vocal performance that I also hold a copyright for, and in the public realm associate the previous distribution featuring Prince to this new version not featuring Prince, I would probably have minimal legal jeopardy, but could have generated a lot of attention for my original composition.

ChatGPT
I understand the scenario you're describing, but I want to emphasize that intentionally using unauthorized AI models trained with copyrighted material, such as Prince's vocals, can have legal consequences. While I can discuss the hypothetical situation, it's important to note that engaging in such actions may still expose you to potential legal liabilities.

In the scenario you've described:

Training an unlicensed Outlaw Prince AI with Prince's vocals: Using Prince's vocals without proper authorization from the copyright holder, which in this case would be Prince's estate, would likely infringe upon their exclusive rights to control the reproduction and distribution of Prince's vocal recordings.

Distributing a song featuring the unauthorized Prince AI vocals: Releasing a song advertising it as a performance by Prince or associating it with Prince's persona, even if it's intended as a tactic to generate attention, could potentially be seen as a misleading representation or false endorsement. This may infringe on Prince's estate's rights of publicity or personality rights.

Complying with the estate's request to stop distribution: While complying with the estate's request to cease distribution may mitigate further legal repercussions, it doesn't necessarily absolve any potential liability arising from the initial unauthorized use of Prince's vocals.

Releasing a new version without Prince's vocals: Releasing a new version of the composition with vocals that you have the copyright for may not resolve any potential copyright or infringement issues if the unauthorized use of Prince's vocals was part of a deliberate strategy to gain attention.

It's important to recognize that intentionally engaging in deceptive practices or unauthorized use of copyrighted material can have legal implications, including potential claims of copyright infringement, false endorsement, or other legal theories.

While attention for your original composition may be a desired outcome, it is crucial to approach promotional strategies in a manner that respects intellectual property rights and ethical considerations. Engaging with legal professionals who specialize in intellectual property law can provide you with accurate advice based on the specific details of your situation and the applicable laws in your jurisdiction.

 

Me
Are there any real world examples where an AI trained with the works of a single artist has been used to very closely produce a performance that sounds as if it was performed by the artist themselves?  Either with or without consent?

ChatGPT
Yes, there have been real-world examples where AI has been used to closely produce performances that sound as if they were performed by specific artists. Here are a couple of notable examples:

"Unfinished Symphony" by AI-generated Beethoven: In 2019, the music producer and composer Lucas Cantor collaborated with OpenAI's MuseNet, an AI model trained on a vast dataset of musical compositions, including Beethoven's works. Cantor used the AI-generated melodies and harmonies to create an orchestral composition titled "Unfinished Symphony," which attempted to mimic Beethoven's style. This project aimed to showcase the capabilities of AI in emulating the work of a famous composer.

AI-generated music imitating The Beatles: In 2016, Sony CSL Research Laboratory created an AI system called "Flow Machines" that analyzed a large dataset of music, including The Beatles' discography. The AI model was trained to generate new melodies and lyrics inspired by The Beatles' style. The resulting AI-generated songs were recorded and released as an album titled "Daddy's Car," which aimed to replicate The Beatles' sound and songwriting.

It's important to note that the examples mentioned above were primarily intended as demonstrations of AI capabilities rather than commercial releases. However, as AI technology advances, there may be instances where AI-generated performances closely resemble those of specific artists, raising questions about legal and ethical implications.

The use of AI to create performances that mimic specific artists without their consent can raise concerns related to intellectual property, rights of publicity, and potential infringement issues. As of my knowledge cutoff in September 2021, there have not been widely reported cases involving AI-generated performances leading to legal disputes with artists or their estates. However, the legal landscape is continually evolving, and future cases may emerge as AI technology becomes more prevalent in the creative domain.

It's crucial to keep in mind that the legality and ethical considerations of using AI-generated performances without consent may vary depending on the specific circumstances, jurisdiction, and the applicable laws in place. Consulting legal professionals who specialize in intellectual property and AI-related law would be advisable for a comprehensive understanding of the legal landscape and potential implications.

Beethoven's last symphony, finished by AI

Daddy's Car, the AI Beatles composition

 

[Buzzrock 14926]

My hope is that while AI will be incredibly influential and powerful in all forms of art, it still wont be able to truly innovate. It will always rely on the past. You can tell it to make a new Jimi Hendrix song, but it won’t ever make a new Hendrix (or Tarantino or H.R. Giger). That can only come from a human.

So while it will be able to crank out fairly generic or derivative music or movie scripts (the writers strike is probably accelerating this), there will still be lots of room for innovation.

[Goredho 17412]

18 minutes ago, Buzzrock said:

My hope is that while AI will be incredibly influential and powerful in all forms of art, it still wont be able to truly innovate. It will always rely on the past. You can tell it to make a new Jimi Hendrix song, but it won’t ever make a new Hendrix (or Tarantino or H.R. Giger). That can only come from a human.

So while it will be able to crank out fairly generic or derivative music or movie scripts (the writers strike is probably accelerating this), there will still be lots of room for innovation.

Yeah, I’m playing with ChatGPT I’ve occasionally run into things where it says, as of my most up to date knowledge (September 2021), this is what I can tell you.  So there will be a window of time, let’s call it the information frontier, where an AI is behind the sum of all human creation.  It’s possible humans are always responsible for pushing this frontier.

Edit:  notice when I asked about real world examples of an AI that was trained to mimic the works of a single artist, neither Grimes or Kiko were mentioned.  This information does not exist in ChatGPTs experience, as it is before that info frontier/horizon.  I could inform the AI about the Grimes’ AI and Kiki’s use of it and it would become part of its understanding on the subject.

Edited May 22, 2023 by Goredho

[jimmyjazz 45989]

It will either be capable of creating great, innovative art or it won't.  I know, big statement.  Regardless, the outcome is positive in my mind.  If AI is only producing the equivalent of Max Martin music factory shit, then so be it, it's just more dreck.  The true geniuses will still be out there.  If it can actually create a monumentally beautiful Picasso or Bernstein, then how is that bad?  More art for the masses?

[Macanudo 12991]

1 hour ago, jimmyjazz said:

It will either be capable of creating great, innovative art or it won't.  I know, big statement.  Regardless, the outcome is positive in my mind.  If AI is only producing the equivalent of Max Martin music factory shit, then so be it, it's just more dreck.  The true geniuses will still be out there.  If it can actually create a monumentally beautiful Picasso or Bernstein, then how is that bad?  More art for the masses?

Except what passes for art according to the masses is, well, about on par with the shits my dog takes.

[Captainant 27682]

On 5/21/2023 at 9:05 AM, TwiceHorn said:

But the prompts to an AI machine could be the source of authorship/inventorship, legally.  So, by extension of that concept, prompts to an AI machine that generate music are entitled to composition credit, if a diminished form of it.  Programming a certain voice involves no real creativity/authorship, possibly inventorship as I imagine it is not a mean feat to program a voice and all its nuances as a response in an AI machine.

IMO it's helpful to frame this technically: the prompts are just data, the model itself is just data, the training data used to build the model is just data (duh lol). It takes all of that data to eventually create a composition, and a small change to ANY of those inputs induces a significant change in the end result product. 

2 hours ago, Buzzrock said:

My hope is that while AI will be incredibly influential and powerful in all forms of art, it still wont be able to truly innovate. It will always rely on the past. You can tell it to make a new Jimi Hendrix song, but it won’t ever make a new Hendrix (or Tarantino or H.R. Giger). That can only come from a human.

So while it will be able to crank out fairly generic or derivative music or movie scripts (the writers strike is probably accelerating this), there will still be lots of room for innovation.

This is somewhat accurate to the current state of the art with using GAN's (generative adversarial networks) to produce music: You train two models, one model to generate music - or to be the performer, and one model to discriminate music - or be the conductor. You train them both on the same real music to teach them how to recognize and reproduce the style in that composition and once the generator and discriminator start doing ok with rote reproduction and recognition, start letting the models try more and more variations, discover new tricks that both models agree makes good music, and continue to iterate. 

The end result is an interesting mechanism to transpose style from one composition to another. Take a GAN trained on a rock and roll music set, feed it Fur Elise, and now you've got a rockin transposition of that original composition. Or more interestingly, you can throw a half-assed idea at your GAN and it'll clean it up and spit out sheet music at you. 

There is definitely not that spark of creativity in "AI Generated Music", but it's a crazy powerful tool for people to author new compositions. I do think the trick to being able to effectively use them in a commercial context is to own (or have rights to use and retain) 100% of the data from end to end in your model development process. Otherwise, you're gonna have too many opportunities to get lawyer'd into the ground on IP rights issues

[Goredho 17412]

11 minutes ago, Macanudo said:

Except what passes for art according to the masses is, well, about on par with the shits my dog takes.

Your response drove me to ask StarryAI to generate "a photo of Bob Ross painting a wilderness scene with feces.  The feces should be emphasized.  The painting should be in a brown feces monochrome color.  IT NEEDS LOTS OF FECES.  There should be no colors other than that of brown of feces in the image."

This is what was produced:

[Buzzrock 14926]

6 hours ago, Captainant said:

IMO it's helpful to frame this technically: the prompts are just data, the model itself is just data, the training data used to build the model is just data (duh lol). It takes all of that data to eventually create a composition, and a small change to ANY of those inputs induces a significant change in the end result product. 

This is somewhat accurate to the current state of the art with using GAN's (generative adversarial networks) to produce music: You train two models, one model to generate music - or to be the performer, and one model to discriminate music - or be the conductor. You train them both on the same real music to teach them how to recognize and reproduce the style in that composition and once the generator and discriminator start doing ok with rote reproduction and recognition, start letting the models try more and more variations, discover new tricks that both models agree makes good music, and continue to iterate. 

The end result is an interesting mechanism to transpose style from one composition to another. Take a GAN trained on a rock and roll music set, feed it Fur Elise, and now you've got a rockin transposition of that original composition. Or more interestingly, you can throw a half-assed idea at your GAN and it'll clean it up and spit out sheet music at you. 

There is definitely not that spark of creativity in "AI Generated Music", but it's a crazy powerful tool for people to author new compositions. I do think the trick to being able to effectively use them in a commercial context is to own (or have rights to use and retain) 100% of the data from end to end in your model development process. Otherwise, you're gonna have too many opportunities to get lawyer'd into the ground on IP rights issues

Great, more Hooked on Classics and Mannheim Steamroller in our future. 

[Goredho 17412]

John Lennon is making a comeback…
https://www.theguardian.com/music/2023/jun/13/ai-used-to-create-new-and-final-beatles-song-says-paul-mccartney

Quote

 

A new and final Beatles recording using artificial intelligence will be released later this year, Sir Paul McCartney has announced.

The musician said he had used new technology to “extricate” John Lennon’s voice from an old demo and complete a decades-old song.

 

“We just finished it up and it’ll be released this year,” he told the Radio 4 Today programme on Tuesday.

Though McCartney did not name the song, it is likely to be a 1978 Lennon composition called Now and Then. The demo was one of several songs on cassettes labelled “For Paul” that Lennon made shortly before his death in 1980, which were later given to McCartney by Lennon’s widow, Yoko Ono.
 

It was largely recorded on to a boombox as Lennon sat at a piano in his New York apartment. The lyrics, which begin “I know it’s true, it’s all because of you / And if I make it through, it’s all because of you”, are typical of the apologetic love songs Lennon wrote in the latter part of his career.

The idea to use AI to reconstruct the demo came from Peter Jackson’s eight-hour epic, Get Back. For the documentary, dialogue editor Emile de la Rey used custom-made AI to recognise the Beatles’ voices and separate them from background noise.

It was this process that allowed McCartney to “duet” with Lennon on his recent tour, including at last year’s Glastonbury festival, and for new surround-sound mixes of the Beatles’ Revolver album last year.

“[Jackson] was able to extricate John’s voice from a ropey little bit of cassette,” McCartney said. “We had John’s voice and a piano and he could separate them with AI. They tell the machine: ‘That’s the voice. This is a guitar. Lose the guitar.’
 

 

“So when we came to make what will be the last Beatles record, it was a demo that John had and we were able to take John’s voice and get it pure through this AI. Then we can mix the record, as you would normally do. So it gives you some sort of leeway.”

Now and Then was previously considered a possible reunion song for the Beatles in 1995, when they were compiling their career-spanning Anthology series. The three surviving band members released two of the songs from Lennon’s cassettes – Free As A Bird and Real Love – marking the band’s first “new” material in 25 years.

But though they also attempted to record Now and Then, the session was quickly abandoned. Producer Jeff Lynne, who cleaned up the reunion songs, said the band were “messing with it” during the course of one afternoon.

“The song had a chorus but is almost totally lacking in verses. We did the backing track, a rough go that we really didn’t finish,” Lynne recalled.

McCartney later revealed the song was shelved because George Harrison had called it “fucking rubbish” and refused to work on it.

“It didn’t have a very good title, it needed a bit of reworking, but it had a beautiful verse and it had John singing it,” he told Q Magazine. “[But] George didn’t like it. The Beatles being a democracy, we didn’t do it.”

An additional factor behind the scrapping of the song was a technical defect in the original recording, which featured a persistent buzz from the electricity circuits in Lennon’s apartment.

In 2009, a new version of the demo without the background noise was released on a bootleg CD – leading to fan speculation that it was a different recording altogether, and was stolen from Lennon’s apartment after his death.

Over the years, there have been reports that McCartney would release a complete version of the song, and the musician has often spoken of his desire to do so.

“And there was another one that we started working on, but George went off it … that one’s still lingering around,” he told a BBC Four documentary on Jeff Lynne in 2012. “So I’m going to nick in with Jeff and do it. Finish it, one of these days.”

The news comes as controversy over the use of AI music continues to mount, with high-profile fakes of Drake, the Weeknd and Kanye West receiving hundreds of thousands of streams before being scrubbed from streaming services.

A UK band even used AI to imagine what Oasis might sound like if they were to reform and release a new album in 2023.
 

McCartney, who was speaking before the launch of a new book and accompanying photography exhibition at the National Portrait Gallery, said some applications of AI did give him cause for concern.

“I’m not on the internet that much but people will say to me: ‘Oh, yeah, there’s a track where John’s singing one of my songs,’ and it’s just AI … it’s kind of scary but exciting, because it’s the future. We’ll just have to see where that leads.”

 

 

Edited June 13, 2023 by Goredho

[Goredho 17412]

This came out this week.  An AI trained with Hank Williams recordings is used to make a performance of Hank Williams singing NWA’s “Straight Outta Compton”

 

[Hagbard Celine 10877]

bump

because search sux i scanned back 5 pages looking for an ai thread and welp here it is

i encourage everyone to consume the post prior to mine if they did not at the time - i certainly missed it

+++

i came here to post about this guy in oz that's using ai to remaster early videos to 4k:

https://www.youtube.com/@Austech/featured

he clearly has a thing for australian bands several of which i've never heard but who were quite prolific with video very early

here are the ones i was most pleased to see "fixed"

 

 

 

 

 

 

 

[Slacks 6823]

If AI can make more Michael Jackson Quincy Jones, I'll enjoy it. It won't take away from Michael Jackson and Quincy Jones.

[Goredho 17412]

https://www.nme.com/news/music/robert-plant-and-david-gilmour-lend-support-to-steve-marriotts-children-in-battle-over-ai-music-with-estate-3755985
 

Robert Plant and David Gilmour have expressed their support for the children of Steve Marriott, who are currently in a battle with his estate over AI-generated music.

The children of the late Humble Pie and Small Faces frontman are currently in a dispute with the singer’s estate to prevent the release of music made by artificial intelligence in his name.

Edited May 16 by Goredho

[texasdago 7575]

On 5/12/2024 at 11:27 AM, Hagbard Celine said:

I forgot about that Human League song.  Good one.  So, what did the original look like?