Suspect Eminem himself would have a more nuanced take on this than the ‘of course this is okay’ or ‘this is never okay’ responses it’s getting right here.
Number 1: Hip hop’s built (to at least as great an extent as EDM) on sampling. And while the degree of respect for clearing and getting authorization to use samples has varied over the years and between artists, as a reference point, Eminem’s breakthrough ‘My Name Is’ uses a Labi Siffre sample as its hook, and Eminem altered the lyrics and content when requested to make changes by Siffre (who is gay, and did not want to be associated with homophobic slurs, for example) in order to get clearance to use it. So there’s evidence there is some basic respect for the artist you’re borrowing from over and above just doing the minimum the law allows or asserting your rights over them. So you know, it might have been nice for Guetta to call and let him know, you know?
2: he’s also part of a culture of musical insults and attacks and rap battle dissing, where throwing out names of other artists and making jokes about them (and worse) is part of the game (see my tongue-in-cheek sibling comment ‘nobody listens to techno’, which is an Eminem lyric about Moby of all people). So he’s more likely to just drop a mean word about David Guetta in a track in the future and call it even than to take him to court. Which would have the advantage of giving Guetta an awesome sample he could drop into future sets, and so the whole cycle completes.
I think if the venue is paying ASCAP/BMI, then the DJ can play a recording from that entire catalog. That probably doesn’t apply to an AI-generated track, but it’s how recording artists are generally compensated for appearing in a DJ set.
If he used an actual Eminem track or lyric he would be covered by venue ASCAP licensing. As it is he used a clip that he made himself using an online tool based on a lyric that OpenAI randomly generated, that means it’s probably not even copyrightable, so has no rights issues (unless maybe he breached the ToS of any of the services he used to generate it?) and is certainly not sharing in commercial proceeds with the artist impersonated.
I am not sure what gave you the impression that he didn’t call. I don’t think there’s been confirmation either way. I know people aren’t always familiar with Guetta or EDM, but it’s important to know how prolific and genre blending he and his music is. I would give him the benefit of the doubt. He’s at the top of the list of artists who would do this intentionally, but respectfully to cause a discussion. Guetta has many song collaborations with mainstream artists in R&B and Hip Hop. His full songwriting credits list is huge but just looking at the credits page for who he’s done music videos with should give you a clear understanding of him. He needs to show respect to the artist, but he doesn’t need to defend that he did publicly.
A YouTuber called "30 Hertz" has been making several fake Eminem songs (+other artists) for 2 years now, in very good quality.
An example: https://www.youtube.com/watch?v=lMunOszEQHE
This isn't going to be released commercially, as a very successful artist, he knows that better than most. No record label would release songs without clearing things with the relevant artist. Many artists and DJs don't release live sets/remixes because it would be impossible to clear all the samples, I don't see why this would be any different.
David Guetta makes tons of money just performing live, so "not going to be release commercially" feels like a cop-out... But then again, that's sort of how DJing works anyway. Would he have had to pay some sort of royalties had he just remixed one of Eminem's actual works on stage?
In which country is that true? In the UK for example venues need to be registered with the PRS (Performing Rights Society) but their fees are relatively low and the funds they collect are distributed based on surveys with the venues - once every year or two someone will come to the venue and ask them to write down the top 20 or so artists / tracks that they play. It's not very targeted or accurate.
Sorry, didn't mean to suggest it was target specifically to Eminem, just that some of the money would get to him via the general performance license. As far as I understand PRS is a common model.
Something like that, lyrics about other artists are often cleared too. There’s a mutual understanding in certain genres, but there’s the infamous example between Taylor Swift and West where the conversation was leaked because West mislead Swift. The candidness of it and the beginning of the line Taylor was willing to let slide was interesting. That wasn’t even AI or a sample, just a reference in a demeaning line.
On the other hand, you might be able to clear royalties quicker, so long as you emphasize its AI with permission, now nobody has to spend money on studio time recording new vocals for some one off DJ remix.
This seems very close to remixing or covering music, which already happens on a big scale, with royalties in place. If they can clearly recognize someone's voice/style, then they would want a piece of the pie. Outside the law, music industry still has a lot of power, and platforms like Spotify are not going to risk those relationships.
Sure but that existed before, you could always ask somebody else to write you a song in the style of Eminem to use in your own song, the difference is how easy it became.
For somebody like David Guetta, it always has been financially possible.
Does Supabase owe him royalties when they use that same model that trained on Eminem’s lyrics to make a chat bot that can interact with their support documentation?
Universal tried to make the same argument with Sony and the VCR. They wanted a royalty payment on every sale. They won their case!
It took the Supreme Court to overrule, where they created the legal doctrine of “commercially significant non-infringing use” as part of determining fair use. They looked at the VHS rental market and the numerous other industries that were dependent on this technology. The courts do not want to be in a position where a ruling can have a profound impact on commerce especially when balanced with extending copyright claims at the detriment to freedoms of speech.
This is why legal experts keep saying the model and the outputs will be considered different and why the model itself is fair use and why the risk-averse lawyers at Microsoft are getting behind these tools… and even why they are letting people build on top of them before they were ready for prime time, to flood the market with non-infringing uses!
While I think the case you're presenting is very likely, I don't think Sony v Universal applies too significantly. The VCR was demonstrably not in any way based on Universal's IP - it was only a tool that could be used to create copies of Universal's IP.
In their capacity as tools, LLMs probably would fall under a similar model. However, the LLM itself is substantially based on the IP of all of the creators of the data used in its training set. This may or may not be granted fair use status, or it may or may not even be seen as infringing on the copyright of those works at all - but either way, it is not covered by that case as a legal question (at least in my own non-lawyer interpretation).
Sony v Universal established a very important legal doctrine with regards to "commercially significant non-infringing use". You can take my word for it, you can go an do your own research, you can confirm with an IP lawyer, or you can wait for the the court's opinion.
Or I guess you can give me a little bit of time to go and help you do some of your own research, which I will do right now, so just hold on a bit!
I'm not trying to trick people or win some hypothetical argument, I'm trying to help people see how the courts will consider these issues and how and why we should agree with their rulings that these tools are fair use of copyright protected works!
I love copyright, probably more than most people on these forums, but the liability needs to be on the people using these tools. Supabase Clippy is fantastic and they should not bear any costs, even implicitly through OpenAI paying royalties. Someone who releases a "Sarah Anderson Cartoon Maker" tool using Stable Diffusion should still be found to wrong Sarah Anderson's common law right to publicity just as someone who releases a "Vacation Photo Background Cleaner-Upper" tool using Stable Diffusion should not bear any costs, even implicitly through StabilityAI paying royalties to Sarah Anderson.
These must be considered in their capacities as tools regardless of how they were made and for reasons foundational to the law itself: How can you prove that I used a given tool such as Stable Diffusion or the "Vacation Photo Background Cleaner-Upper" tool without concrete evidence of the tool, such as the presence of the software on my laptop? Can law enforcement get a warrant to search based on literally no visual evidence that Sarah Anderson's works were somehow used in the training process of a tool that I have on my private property?
Edit:
---
In the Texas Law Review in March, 2021, Mark Lemley, a Stanford law professor, and Bryan Casey, then a lecturer in law at Stanford, posed a question: "Will copyright law allow robots to learn?" They argue that, at least in the United States, it should.
"[Machine learning] systems should generally be able to use databases for training, whether or not the contents of that database are copyrighted," they wrote, adding that copyright law isn't the right tool to regulate abuses.
But when it comes to the output of these models – the code suggestions automatically made by the likes of Copilot – the potential for the copyright claim proposed by Butterick looks stronger.
"I actually think there's a decent chance there is a good copyright claim," said Tyler Ochoa, a professor in the law department at Santa Clara University in California, in a phone interview with The Register.
In terms of the ingestion of publicly accessible code, Ochoa said, there may be software license violations but that's probably protected by fair use. While there hasn't been a lot of litigation about that, a number of scholars have taken that position and he said he's inclined to agree.
Both Lemley and Ochoa state that the models themselves are probably protected by fair use. Meaning, it is perfectly fine for OpenAI to train their models on publicly accessible copyright protected works without asking for permission or having to pay any royalties or having to adhere to any of the terms of the license.
They are also free to distribute this tool and to charge for people to use this tool.
What Ochoa is saying about there being a good chance of there being a copyright claim is that this tool doesn't absolve the users of the tool from copyright violation. The liability is on the person using the tool regardless of the tool being used. It's not the intent that matters with copyright, it's that you ended up publishing something that looks enough like someone else's picture that twelve people consider it to be not too different than a simple photocopy.
Now, this can still be a problem for Copilot because an engineer's company might not want to be injecting a lot of copyright protected code into their products, but for the most part the outputs from Copilot have been non-infringing. That it sometimes produces infringing code does not matter to anyone other than the person using Copilot.
Ochoa goes on in detail about what is and isn't covered by copyright with regards to code, which is one of the things that gives me confidence to use Copilot and know that I'm not putting myself at risk:
But in terms of where Copilot may be vulnerable to a copyright claim, Ochoa believes LLMs that output source code – more so than models that generate images – are likely to echo training data. That may be problematic for GitHub.
"When you're trying to output code, source code, I think you have a very high likelihood that the code that you output is going to look like one or more of the inputs, because the whole point of code is to achieve something functional," he said. "Once something works well, lots of other people are going to repeat it."
Ochoa argues the output is likely to be the same as the training data for one of two reasons: "One is there's only one good way to do it. And the other is [you're] copying basically an open source solution.
"If there's only one good way to do it, OK, then that's probably not eligible for copyright. But chances are that there's just a lot of code in [the training data] that has used the same open source solution, and that the output is going to look very similar to that. And that's just copying."
In other words, the model may suggest code to solve a problem for which there's only really one practical solution, or it's copying from someone's open source that does the same thing. In either case, that's probably because a lot of people have used the same code, and that shows up a lot in the training data, leading to the assistant regurgitating it.
So in practice it is pretty easy to tell that Copilot is spitting out purely functional suggestions basically all of the time as there isn't really any other way to wire up a unit test or call a specific API.
Ironically, if Copilot gets better at "software architecture" then it starts to cross over into the expressive parts of software that are indeed covered by copyright, meaning these issues of liability become harder to discern to the end user and enough of a problem that GitHub would want to figure out attribution or somehow "clear" the suggestions for the user.
Edit: after the text you added, I believe we are essentially in agreement. IF it is accepted that the way GPT3 was created is a fair use of the works in the training data, THEN I fully agree that (1) OpenAI has every right to sell it even if (2) some uses of it would still constitute copyright infringement, since (3) only specific users would be liable for copyright infringement in their uses.
Where we differ is in how certain we are that the IF is true. I for one believe there is a good chance that the training of an LLM on copyrighted works does infringe on the copyright of those works (if no other exceptions apply, such as the LLM being trained only for academic research purposes, of course).
My original response:
Where Sony v Universal definitely applies though is when evaluating whether OpenAI's selling of GPT3 to others who then use it to create copyright-infringing works would make OpenAI liable for contributory infringement. Here, the similarities are crystal clear, and the conclusion is simple: since there clearly exist non-infringing uses of GPT3 (such as Supabase Clippy), OpenAI is fully in the clear to sell GPT3, just as much as Sony was for selling the VCR.
However, this assumes that OpenAI has the rights to the IP of GPT3 itself in the first place, which is a prerequisite to them being allowed to sell it at all. Sony certainly had the rights to the IP of the VCR - Universal never claimed that the VCR was a derivative work of their movies.
Essentially, in Sony v Universal, Universal was claiming (1) that Sony was liable for contributory infringement, since (2) all customers of Sony who used it to record and then playback a Universal show were guilty of copyright infringement. The court established that (2) was in fact fair use, and from there automatically (1) become false, since now there was an established legal way of using the Sony product.
But, in a hypothetical OpenAI v Universal, Universal could plausibly claim that (1) OpenAI is liable for copyright infringement directly, since they are distributing GPT3 , (2) which is a derived work of Universal's IP used in the training set of GPT3.
I honestly think you're more certain because you want the courts to rule in a certain manner.
I'm more certain because I'm thinking about this separately from my opinions as to how the courts will rule. Yes, I will just so happen to agree with that ruling because I happen to agree with the logic and knowledge contained in our legal process. I agree that the existing legal doctrines already capture the spirit of what we are asking them to judge. I agree that the statutory law, case law and doctrine that informs their judgement will successfully balance both the limited rights of copyright holders and the natural rights of a public to unburdened access to the arts, knowledge and information. I agree with their process of balancing the potential impact on existing commercial practice with the potential impact on new forms of commercially significant non-infringing practices.
Some things in copyright might just seem unfair, like the case of Baker v Selden:
In 1859, Charles Selden obtained copyright in a book he wrote called Selden's Condensed Ledger, or Book-keeping Simplified. In it the book described an improved system of book-keeping. The books contained about twenty pages of primarily book-keeping forms and only about 650 words. In addition, the books contained examples and an introduction. In the following years Selden made several other books, improving on the initial system. In total, Selden wrote six books, though, evidence suggests that they were really six editions of the same book.
Selden, however, was unsuccessful in selling his books. He originally believed he could sell his system to several counties and the United States Department of the Treasury. Those sales never happened. Selden was forced to assign his interest—an interest that apparently was returned to his wife after his death in 1871.
In 1867, W.C.M. Baker produced a book describing a very similar system. Unlike Selden, Baker was more successful at selling his book–selling it to some 40 counties within five years.
Selden's widow, Elizabeth Selden, hired an attorney, Samuel S. Fisher, a former Commissioner of Patents. In 1872, Fisher filed suit against Baker for copyright infringement.
The poor old widow lost. Boohoo. But this was a just ruling!
I'm not sure that the people who think that ChatGPT is guilty of copyright infringement are thinking about the issue in a balanced manner. Luckily our courts probably will!
One strategy that the defense could use to lower their risk profile is to allow open access to their models and allow an entire ecosystem of commercially significant non-infringing uses to blossom because they are aware of how the courts will be influenced based on existing statutory and legal doctrine...
OpenAI will 100% need to start paying royalties. For instance it knows a whole lot of things that are not publicly available and locked away behind science journal paywalls.
While the machine learning community has begun to discuss the issue of the legality of training models on copyright data, there is little acknowledgment of the fact that the processing and distribution of data owned by others may also be a violation of copyright law. As a step in that direction, we discuss the reasons we believe that our use of copyright data is in compliance with US copyright law.
Under pre (1984) (and affirmed in subsequent rulings such as aff (2013); Google (2015)), non-commercial, not-for-profit use of copyright media is preemptively fair use. Additionally, our use is transformative, in the sense that the original form of the data is ineffective for our purposes and our form of the data is ineffective for the purposes of the original documents. Although we use the full text of copyright works, this is not necessarily disqualifying when the full work is necessary (ful, 2003). In our case, the long-term dependencies in natural language require that the full text be used in order to produce the best results (Dai et al., 2019; Rae et al., 2019; Henighan et al., 2020; Liu et al., 2018).
Copyright law varies by country, and there may be additional restrictions on some of these works in particular jurisdictions. To enable easier compliance with local laws, the Pile reproduction code is available and can be used to exclude certain components of the Pile which are inappropriate for the user. Unfortunately, we do not have the metadata necessary to determine exactly which texts are copyrighted, and so this can only be undertaken at the component level. Thus, this should be be taken to be a heuristic rather than a precise determination.
1984. Sony corp. of america v. universal city studios, inc.
2003. Kelly v. arriba soft corp.
2013. Righthaven llc v. hoehn.
Chat bots are not in competition with Eminem in his industry. Whereas other music artists are.
Its entirely fair game for subjects in a given field to protest when an AI is trained on their work and then is used to power competition to them without any form of remuneration or licensing to compensate them.
Great, so we agree that it is fair use for OpenAI to train on copyrighted works as long as the person using the model does so in non-infringing ways and that the liability rests with the person using the tool.
Logically, the model must be considered in fair use if we are to allow Supabase Clippy. This is how the courts will see it!
None of his discography is being used verbatim. The UK has already determined AI training is explicitly protected use, and does not require rights licenses or royalties.
Some countries will try to hamstring AI development by saddling it with perversions of copyright use, and those countries will increasingly fall behind technologically and economically.
Guetta is not saying it's the real deal. The creativity, is his usage of the tools available. He could have hired an impersonator, it's the same thing (in my opinion).
But that’s in a commercial setting, this is during a musical performance. The audience was not tricked into showing up to an Eminem performance and were not tricked into thinking that Eminem was on stage at any point. This is protected speech.
Sure, but in the Bette Midler case, I was talking about a literal television commercial:
Midler pursued a common law judgment against Ford for using her distinctive voice without her authorization. The appellate court pondered the question of whether or not an artist's voice is a distinctive personal feature over which a person has controlling rights from appropriation. Midler was not seeking damages for copyright infringement of the song itself, but rather for the use of her voice which she claimed was distinctive of her person as a singer. The recognition of Midler's voice in the commercial was found to be the intentional motivation and a major feature of the commercial.
This is a common law right to publicity. This is our common law right to be in control of our likeness.
David Guetta is being paid but he is not being paid by the promoters or the audience because he says he is Eminem so there is no reasonable claim to a violation of Eminem's common law right to publicity.
Whereas with Midler, the commercial production crew was seen by the jury to be paying an artist to convince an audience that Bette Midler was endorsing the product.
Chiming in here, even if he was not being paid for performing this song here, the tweet and publicity still serve to market his work, right? How is that not commercial? I mean, this is to sell more tickets in the future, one would assume.
According to a quick google search, even cover bands performing live are required to pay royalties to the original artists whose songs they're performing.
Maybe there's some loophole when using an artist's likeness while changing the actual music and lyrics, I don't know.
But "protected speech" isn't meant to be a way to make money off of other people's work.
Curious what part of "Midler v. Ford Motor Co" you think wouldn't apply to a musical performance.
Do I need to ask permission to impersonate Eminem during a comedy performance? How about during a musical performance?
The answer is no. This is fair use 101.
If you cover a song of his while doing an impersonation then this is covered by the performance fees the venue you’re playing at pays every year to ASCAP and BMI.
Satire protects you only if you needed the thing imitated to make your point [and you have $$$ for lawyers to argue that]
The example I like (even though Weird Al always obtains permission so doesn't need Free Speech protection) is "Smells Like Nirvana" compared to "Fat".
Smells Like Nirvana is satire. The thing criticised is Nirvana's "Smells Like Teen Spirit", and the wider Grunge phenomenon it exemplifies. Al's song talks about how Kurt's lyrics are mumbled and hard to understand, "It's hard to bargle nawdle zouss / With all these marbles in my mouth", and about how it's too loud so it'll annoy your mom and dad, "We're so loud and incoherent / Boy, this oughta bug your parents". This can't work if Al uses Mariah Carey's "All I Want For Christmas" instead of Teen Spirit.
"Fat" isn't satire. Michael Jackson wasn't obese or even overweight, and his song isn't causing people to eat too much junk food, Al's song is just using the outline structure of Jackson's "Bad" because that fits and his new song is funny. If his lyrics had fitted better to Rick Astley's "Never Gonna Give You Up" that would have been fine.
No, you’re not correct. Cover bands don’t need to ask permission. The venues they play at should be writing down the songs they’re playing and submitting them to PROs, but the bands don’t need to do anything.
> originals that the cover band did in the style of anyone else
Interesting. I guess you consider the use of Eminem’s voice to be an ‘in the style of’. I think, being trained on Eminem songs, it’s more like a sample.
Fair use is relevant for parodies that are not subject to copyright or royalties so it is indeed relevant, but you're right, the issue with Bette Midler is the common law right of publicity, of which David Guetta is not in the wrong because no one was tricked into thinking that he was Eminem. Even if David Guetta's DJ sets became nothing but Eminem songs it would be fine because he would still not be tricking anyone into thinking that he was Eminem.
> According to a quick google search, even cover bands performing live are required to pay royalties to the original artists whose songs they're performing.
It's more than that. If you were part of a band and now you are solo (plenty of cases), you need to pay your old band if you play an old song you were part of (unless you own 100% of the lyrics and music rights, which is rare).
By the way, I am totalyl OK with this deal/agreement/way of doing things.
I believe David Guetta that he was really just playing around with these cool tools here (who isn't blown away by them?). But I guess it won't go unnoticed by the teams and investors of uberduck and other AI startups that this is the perfect guerilla marketing stunt.
Gotta admit it, I am guilty too, never heard of uberduck before and caught myself creating an account and browsing their pricing site today.
Maybe I am late to recognize that, but with recent developments I get the feeling that AI and machine learning are really getting somewhere now. If it goes forward with the current trajectory, then this could change the world just like radio, television or the internet did.
How so, isn't that exactly what it is doing to an even greater degree? Those who would never have the capability of doing something are now able to do something they only dreamed about.
It is exactly the inverse of previous advancements, in that those already at the top don't get nearly as much benefit as those at the bottom since I will be able to copy your skill no matter how good you are and reproduce it infinitely.
The internet broke down the barriers to content distribution. AI will break down the barriers to content creation.
The future will not be "look I made an AI song that millions of people are listening to!" but will be "AI create and play me a song of my favorite rapper overlaid onto of my favorite techno genre" and you'll just listen to it yourself.
Today, right now, you can have a language model create lyrics to "AI killed the Internet star," a TTS model sing it in the style of The Buggles, and a single channel blind source separation model replace the voice track. Tomorrow, you will just talk to your voice assistant running an action transformer to have it done for you.
Imagine in the future when we will have AI ways of doing things locally on our smartphones, how would this be similar to radio/tv/internet changing everything?
You could use https://vocalremover.org/ to remove music. I've used it for sampling speech out of tv programmes with background audio and the results are immense.
Interesting, hadn’t heard of Tortoise TTS. You can use something like Meta’s Demucs to extract the vocal from a full track in surprisingly good quality too.
But who will those gansta rappers turn to, should some hn-gansta steal there IP, likeness and ID?
We should create a retribution gang, that rolls up with flashing lights and distributes punishment. They could hang out at strip clubs and cocktailbars while not needed.
We could call it pole-ICE..
History producing strange staircase humor..
But these moments, they were what they told me to expect, when the singularity begins to dance through the streets..
Last time I checked, David Guetta was not producing techno. Eventually he spends most of his time with his arm in the air so calling him a dj is a huge stretch too.
The rules for live DJ sets is very different from copyright rules that limit published music. This is the same as when artists use samples of meme audio, which otherwise would be copyrighted. Guetta has a long very large body of work and deserves the benefit of the doubt when it comes to the question of Eminem okaying it. He very clearly wants there to be a discussion about this. Celebrity impersonators for singing have been a thing for a long time. This isn’t about one artist in particular.
I was playing with Tortoise TTS and was genuinely surprised by how good it is with just a few minutes of clean audio. It didn’t take me hours to train or fine tune, the generation step is sort of long, 5-7 minutes for 30 seconds, but it feels really similar to stable diffusion where you do quick test with slow samples and iterations to find a decent seed, and then you let it do a more complete regeneration. It’s zero shot generation that ran on my laptop 2070 max q and i7 10750h. It’s not perfect but it’s believable when layered with music.
I just want an entirely new collection of music by the AI formerly known as Prince so that we can move the lawsuits forward and rule in favor of the new-creator. Derivative works will win. Prince will roll in his grave.
I saw this via a LinkedIn post which said it used ChatGPT and uberduck.ai, the search on there doesn’t seem to work for posts so I can’t find the original source unfortunately, but it would make sense that he used those tools
I'd bet he got more lyrics than that from ChatGPT, but was unable to make Uberduck "rap" something that matches the rhythm in any meaningful capacity. So he ended up using just a short snippet of the output.
Even the real Eminem sounds horrible around that speed (~130 BPM), you're way more likely to hear a Drake mashup in that range (or up-until-recently Kanye). I know these sorts of famous AI voice generators don't have the notion of speed or rhythm, but I'd still bet AI-generated Drake would've been much easier to work with.
ChatGPT can emulate rappers who use versus create new language, metaphors and expressions. It’s a natural impression twice distilled from its origin, a lived experience.
Wow, people seem so ready to be aggro about Guetta or ai or w/e. Disappointing that people can't get past their high school opinions and have some fun.
Check out his sets from Dubai and so on over the pandemic, Ultra, etc.. Dude brings good vibes, relax
It seems similar to the DJ process though, right? Their creation process involves sampling, mixing, curating, etc. so the novelty is in figuring it out and getting it to work with his set.
I’m sure the first person to release a rap record felt proud even though they were copying lots of people who were rapping on the street.
The novelty here is being the first dj to do this for his crowd, not that he invented the process. So it’s exciting to him and the people in the audience.
All of this is true, I think I made the mistake of assuming some domain knowledge here - he's seen as a bit of a joke character by a lot of people in dance music, largely down to his seeming total lack of self-awareness. In some ways it's actually kinda charming, like here he has a child-like joy at having done this, whereas sometimes it's completely tone-deaf, like this: https://www.youtube.com/watch?v=dEI7oX0XxJw
I mean.. there’s a reason that others haven’t really done this as far as I know and that he only used a tiny sample for a couple seconds: the output from AI voice generation generally has a ton of artifacts and isn’t something you’d want to play on a festival sized sound system for more than a few seconds.
So he basically did the only acceptable thing I’d consider doing with this and.. it’s just not that interesting?
It’s certainly your prerogative to not find it interesting and your feelings are valid.
However, I think it’s interesting and can’t wait to see more djs and even new artists using this technique to create some neat art. And of course, it looks like there are tens of thousands of fans in that stadium who seemed kind of interesting.
I don’t think statistical sanity is a good measure for value, especially art, but I don’t think it’s noteworthy that you don’t find it interesting.
I understand that. But he's making a big deal out of using literally the two easiest tools out there to generate... a tiny awkward two second soundbite in a set. I just timed myself and it took me less than two minutes to do the same thing. Might take longer if you need to sign up first.
And this isn't to say that the time it takes to do something is important - the idea of course matters too. But the idea itself isn't original. And the result is just a TINY soundbite.
So it's just straight up not interesting from a production nor listening perspective, especially given the fact there are people out there doing much more creative, involved, and interesting things in the realm of AI assisted production. For example: https://twitter.com/thisiscyclops/status/1619085469198721038...
If David Guetta wasn't a big DJ, no one would care about this.
I don't think this is true any more. At least, I've heard generated audio from the latest models that runs for well over a minute without obvious artifacts.
Elevenlabs, by some ex-googlers: https://beta.elevenlabs.io/ If anything, the results I've heard from this are better than the demos.
And VALL-E by Microsoft, which isn't quite as good, but notable because it can clone real voices with only three seconds of training data:
https://valle-demo.github.io/
The pride isn't from creating it from what I saw, the pride is from finding something that was very popular. He said in an interview that it was a throwaway gag he did and wasn't expecting much from it.
The video reference the common ChatGPT "in the style of" prompt. Is there a list of what "styles" ChatGPT's model has been trained on ? Or is this information not disclosed since it would be an admission that ChatGPT has been trained on copyrighted lyrics, etc. by which is produces these derivative results ?
You can ask ChatGPT which artists it’s able to emulate, and it’ll tell you. I asked for a few sample lyrics from its list and they were all as good as you’d expect.
I think the significance is that something we have been excitedly debating and closely observing for years on this forum is starting to pop up everywhere in the mainstream culture.
Guetta, like most big DJs, definitely uses ghost producers. But he also very obviously knows how to produce as well. And if you are a producer, you're likely technically savvy enough to figure out basic AI tools.
I will also say -- UberDuck and AI TTS in general, when compared to the SURGE of development and tools that's happened on the image/video side of AI, is TERRIBLE.
UberDuck's community specifically seems geared towards kids making memes -- I suspect they just ended up there and didn't design it that way, but wading through the terrible user created models to find ones that work was tiresome.
I tried to get https://coqui.ai/ setup to do similar things, but have not been successful.
Surely this will all explode in the next 18 mo max
Number 1: Hip hop’s built (to at least as great an extent as EDM) on sampling. And while the degree of respect for clearing and getting authorization to use samples has varied over the years and between artists, as a reference point, Eminem’s breakthrough ‘My Name Is’ uses a Labi Siffre sample as its hook, and Eminem altered the lyrics and content when requested to make changes by Siffre (who is gay, and did not want to be associated with homophobic slurs, for example) in order to get clearance to use it. So there’s evidence there is some basic respect for the artist you’re borrowing from over and above just doing the minimum the law allows or asserting your rights over them. So you know, it might have been nice for Guetta to call and let him know, you know?
2: he’s also part of a culture of musical insults and attacks and rap battle dissing, where throwing out names of other artists and making jokes about them (and worse) is part of the game (see my tongue-in-cheek sibling comment ‘nobody listens to techno’, which is an Eminem lyric about Moby of all people). So he’s more likely to just drop a mean word about David Guetta in a track in the future and call it even than to take him to court. Which would have the advantage of giving Guetta an awesome sample he could drop into future sets, and so the whole cycle completes.