Thing is they’re not doing this because they care about the artists. They’re just mad they didn’t do it first. There are already major labels using signed artists to train vocal models. They’d be happy to shovel generated content into our ears is THEY make it and profit from it. It’s not about ethics, or artist rights, at all.
It's funny when all these music people are fighting FOR the big music labels to control the music industry even more. They want "kinda sounds like and reminds me of" to be a copyright violation going forward..... it's crazy.
If the major labels put out a product like suno or udio... I would be a vip for how often I use it, and what I do with it. Just like I am at the other two. I think it's a great idea, and if paying them is what it takes, fine. Paying them is no problem. That said, I don't think they could do the job of training that Suno did without help. The breath of it. It's more than music or tags. That's where most people are getting this wrong. Suno has things in the training data that make it a musical instrument... if you know how to do it. I believe that's because they scanned a century of sheet music as well... among other things. Udio has better fidelity, but theirs is a much simpler training than the Chirp models.
Napster all over again :)))) Record Labels had made a fuss about mp3s.. because it was free.. then the bought Napster, then came Spotify.. now everybody listen digital music.. nobody listens CD or Casette :))) Generative music will prevail eventually... even now, some musicians and producers are using this tools.. b/c it gives creative ideas and precents writer's block.. and everybody with a simple idea can become an artist.. no ne to know notes and sheets and play any insturments and have a great voice :)
But the major labels want to create their own AI music and profit from them. They’re not protecting artists. Spotify wants to create fictional artists for playlists and collect the profits.
Yea and it is no secret they have that one Swedish composer to write all the easy listening piano tracks. Look to find the names they serve as the composers of each individual track and good luck finding a trace of any one of them! These are fake names. They have made 150 million dollars from those tracks and paid him 6.7 million.
@@RideShareRocks I’m glad you can predict the future- you need to market that. Face it, you don’t know who will win. Gen Z is already onto AI music and is rejecting it outright. They have good ears for what is fake and they hate fake. I work with a lot of of Gen Zers and once you lost them, good luck. They are the future of culture and the ones who are ultimately in control because they will refuse to consume it. It will become the equivalent of any product that won’t sell. There was no long term business plan in the Udio or Suno training. Outright rejection of the apparently easy to spot AI songs is something they didn’t factor in, but it could be the fatal blow to a grand experiment that no one asked for.
I think if used in the good way, the AI, especially udio, can help artists that they are in lack of inspiration. Then real artists can go touring and make money with that... AI can't go on tour and for sure the people using AI to create music at the moment are like me that I am so out of tune, that if I sing under the shower, the water will stop and refusing to work again 😂, but sometimes I have some nice ideas that I like to see how would be. And now AI help with that. Isn't that great?
People are also trained on listening to existing music and they also generate music the same way AI does. This is inevitable. These companies are in a state of urge because they are afraid to lose the monopoly they have in the music market. There is no way they can escape from what the future holds, though. They have to adjust, find new ways of earning money from the backs of the musicians or they will not exist in the future. How sad they are not willing to be a part of the future! They still want to direct the trends in the market and direct the people’s taste for music by their huge advantegous hands. But behold, this will change. This will change for the movies and so on too.
It’s very different. These platforms are trained on the actual master recordings, not just being ‘inspired by’ the melodies/lyrics (which would be fine). They are essentially sampling, and usually a human would have to clear that sample with whoever owns the masters.
@@bw2937 I think human mind is also trained on listening to the same masters. You can even not knowingly replicate an album’s or band’s sound when mixing and/or mastering. This discussion is no more than being arbitrary unless there is full scientific research covering human vs AI learning/ interpretation/putting into use with every parameter and aspect compared. Even this will be the first step into making laws accordingly. Generative AI requires a different legislation then the current one. Until then even these lawsuits will result in favor of them, there is no way escaping what is to come next.And AI does not use the masters as is.
I thought. Exactly the same about most of the artists are also trained by this way too in using active hearing and music analysis. This big companies like Sony are just afraid to lose a lot of money. They don’t really care about the artists. For them an artist is just a product as an other one that they just want to sell to win money. I think that AI is a big and good revolution for the civilization. And finally, it isn’t easy to create a good music even with AI. I think that if a person is able to create a wonderful power ballad or other great music or song with AI and that people like that song, then he also deserve to win some money for that and not necessarily these big companies.
@@DigitalAI_Francky I agree, Sony once gave me a strike and wanted to scare me for allegedly using their music. The funny thing was that it was a recording of a Mozart piece of music with my music club, although the sheet music arrangement was already well over 100 years old. Of course, I had proof that it was a copyright-free piece of music and the strike had to be withdrawn.
look at the advancement in sound sample libraries to get a sense of what's coming. Sitting in music theory class in 1977 my instructor (with a PhD) said, "one day these synthesizers will replace musicians'" At that time it was only analog synths. I thought his statement was ludicrous.Look where we're heading. It's really the old story of Frankenstein, isn't it?
The problem is that it does not really matter how the music was made. That is an old way of thinking. We are creating a new way of making music, and the old way stands in the way for progress.
I agree, what is important is not how you create the music, it is if the people like your music or not. If they like, they will buy it and if they don’t they won’t. It is simple. I can buy a music even if I know that I could hear it forever for free in UA-cam but I pay because I considere that the artist deserve to be paid for it.
The AI that UDIO and SUNO uses is based on Musicgen's related technology, which uses a multiband diffusion approach. Musicgen is a open source technology that can create music from prompts as well and it can use a melody from music you upload. I have two tunes where I took the melodies from "Hide and Seek" of Howard Jones and "Desert" by emilie simon (a french pop artist). Believe me, the technique is so well known among the AI world that its not going to matter a hell of beans if the music industry succeeds in stopping SUNO and UDIO, its just going to entice hackers and activists who want to eliminate the record companies cause of the evil behaviors they have exhibited in the past.
As a musician and writer, we must try not to be affected by music that we love in a way that it's to duplicative, and anyone that does write music knows it's hard not to do this. Why should A.I. algo's be any different? Now I'm going to go all rant... These labels have a major fight on their hands. Frankly, they have had it too good for too long, turning good wholesome roots into a teenage "poppy-mill" with roomfuls of writers and pre-teen sex object singers that can't even write their own name. It's about time these moguls poked their heads out of the money pile to have a look around. Competition is scary huh? Boo!
If someone can create a beautiful music or song with the AI, I dont’t really see that it is a problem as long as many people love this music. If I like a music, I will buy it even if I know that I could hear it for free in UA-cam. I like to buy it in ITunes for example because the artist deserve to be paid. Now, I don’t know how much are artists paid from these platforms but at list I pay and I find it normal. The AI will never be in competition with performers artists in live concerts at least and this is not so bad. But it isn’t so easy to create a good music with the AI. Of course you can create very beautiful music too but it takes also a lot of time and try before to get a good result. Everybody want to talk about freedom, so it is also a freedom to be able to create music even just for ourselves too if we want. 😉
I had Udio generate music for me to go along with lyrics I provided with very little prompts otherwise. Out of nowhere, it lifted the main piano riff from Gary Jules' cover of "Mad World" note for note but in a different key. If it had merely shifted a few notes I may not have even realized where it got the sound from. This result surprised me, as I had assumed that the reason other people were getting copyrighted infringement results was due to their prompting, telling the AI to "sound" like something or by providing very similar lyrics. Basically guiding it and/or limiting in ways where it couldn't help but copy something. But there was nothing about my lyrics that matched the original song in words, theme or structure, and it only used the riff when generating the song for me, the melody for my lyrics was entirely different. I gave it pretty wide latitude to create something entirely new and yet it directly copied a significant and defining portion of another song. I've come to realize that generative AI is simply a very splintered plagiarizer. It takes tiny fragments of existing works and rearranges it or modifies it until it's largely unrecognizable. But if it lacks enough variety, the chunks come out bigger and CAN be recognized. It's just a very sophisticated imitation machine. It does not "create" anything, it simply mimics what it ingests. This differs dramatically from what a human does with the content we ingest. We "feel" something and are inspired to create something new from the emotions the music generated, or we understand what the artist was going trying to do and build on that idea and take it somewhere new. AI has no understanding at all of what it generates, it merely identifies patterns and reorganizes them in a way it predicts will be successful. It doesn't even know if what it generated is any good, it only predicts that the odds are good that it might be. I used to think "what's the big deal of AI learning from existing works, humans have to as well, AI just does it much faster." Now I understand that, no, generative AI is not really "learning" anything. You're just giving it more content and patterns to copy and/or imitate. Whereas a human runs the risk of plagiarizing if they're being careless or lazy, AI can ONLY plagiarize.
That's your opinion. Tech is going to evolve as it always does and make people's life more comfortable and give access to use certain things easier. AI draws instead of you, write novels. Am I bothered by it as a hobby novel writer and artist? No. It's all about self expression, the market is there you can choose what you want to see, read, listen. If you are good in what you are doing, with some brain you're gonna get the attention.
Udio is all over the place.. same prompt may end up creating dumb tunes.. or create realistic artistic songs even with real people's voices.. i guess Udio is more trained by actual copyrighted songs.. but it was rather a new release than Suno.. so it is not great than Suno... but it definitely copying real music (i m down with that :)))
@@HERKELMERKEL im not familiar with udio, but i spent hours and hours just to create a song in suno. Never had a feeling its a copy from existing songs.
@@Reealos I am Turkish and i accidentally created exact copy of at least 3-4 of Turkish singers via Udio and yes Suno is not creating a copy.. i am using since discord times.. and i have 100+ song that i have really liked and saved
Crazy Train - ozzy Sweet dreams - Marylin Manson Undead - Hollywood undead These three things have literally almost nothing in common song wise... except a little micro-plagiarizing they do throughout the entire song. Now... Crazy train may have come out with it first, but do we really know if they're the first? The vast majority of rock sounds like rock because people do similar things, micro-plagiarizing in a similar way to the AI generative models. Humans are NOT creative, we are literally the sum of our experiences.
As a fellow musician, I understand your bias. However I think you're overlooking a dramatic difference between Suno and Udio. Udio showed up several months after Suno had completed training on several versions of their models, gradually and with intention homogenizing their output to NOT sound like any existing artist or song. I strongly suspect that their most recent model leans heavily upon licensed studio music. Meanwhile, Udio has not had the time to move beyond the sound of the music initially used for training. Udio is much more cavalier about the music they generate. I think that the music industry is going to have a very hard time proving (by generating songs) that Suno contains any specific song. I've tried hard to trick it. If you try inputting existing lyrics, they never sound like the original song, even with audio input, it doesn't happen. And, Suno makes a pretty strong effort to refuse copyrighted lyrics, too. It will be interesting. But, IMO, you're more likely to violate copyright with your DAW. This is relevant because the only way to even attempt to violate copyright, using Suno, is for the user to intentionally push the program to do so.
I've tried hard to trick it too, and haven't been able to, but I've had friends of mine who have. I have no particular comment on the differences between Suno and Udio as I've tried so many experiments I've lost track of what worked best on each platform.
@@daviddas you can add [chord] annotations to mimick a progression, but Suno probably still won't deliver anything that resembles the original song. In all honesty though, I've occasionally gotten random voices that sounded very familiar... Udio, on the other hand, will gladly imitate the Beatles, etc.
the labels already said that they are in the works of creating their own "models". they just don't want to fall behind again like the whole Napster thing. They don't care about artists.🤣
Ingesting works is what every artist/creator does. This is not copyright violation. Producing new works based on what was ingested over a lifetime is what every artist/creator does. This is not copyright infringement. Forcing/coercing an artist/creator to produce a work that is a forgery or copy of an existing work is illegal. This is what the record companies have done, using exact lyrics and artists' names when using the AI tools. They are the criminals here, not the AI tools.
Spot on ! Humans are doing the exact same ! As a musician my self I can say all musicians start by copying what inspired them in the first place. We generate music this way. They just scared and greedy , they see the writing on the wall.
You're anthropomorphizing. Humans don't function the same way; they retain ideas, feelings, and their passions. The human brain acts as a big filter, whereas an AI's 'brain' merges information extensively - and therein lies the issue. With the right input, AI can produce something very close to the original, albeit with some loss. It's akin to recording songs from the radio onto cassette tapes with hiss and wobble when I was a kid. If I remix two songs from those tapes, is it my creation, or does it remain a remix of existing songs? This concept often eludes people who struggle to grasp how a remix can sound vastly different from its originals, or how a billion songs can contribute to just one remix.
@@High-Tech-Geek are you a federal Judge who is knowledgeable in copyright violation? I think not. It’s really in the hands of those whose job it is to interpret the law. And it could very well go in favor of the copyright holders. Scraping copyrighted music is theft. The law says you must license the music to use it commercially. They didn’t license it. You just don’t know how this will turn out, period.
In the ongoing copyright lawsuits against UDIO and SUNO, we are witnessing a possible future where major music corporations could dominate the AI music scene. If these companies were to claim in the future that since they own the copyrights to the music used to train these AIs, they should also control the entire AI model, then we could see a significant shift in who controls this technology. If these corporations win, SUNO and UDIO will likely fall and be subject to buyouts, further concentrating power with these large corporations. They would then have exclusive rights to AI technologies capable of producing tracks that compete with top artists. This control could severely limit innovation, making it too expensive for new musicians and small businesses to access these tools. The result? A lack of diversity and creativity in music, with the entire AI music industry-or perhaps the entire music industry-falling under the control of a few large companies. We must recognize that allowing a few companies to use copyright laws to monopolize advanced technologies threatens artistic diversity and is against the public interest. Music should be a global treasure. We call on everyone, including policymakers, to focus on this issue to ensure that technological advancements are shared fairly, benefiting all, not just a select few.
As a user of Suno A.I. that used to make music with my keyboard MIDI with FL studio over 20 years ago, it's incredibly fun and powerful tool to realize a vision you have for a song, if you use the custom setting you can write your own songtext and create a mix of genres you are looking for especially if you are very descriptive, sure expect a few hundred retries to get the exact result you want. But this is for sure the future of music creation.
As someone who didn't have the luxury of learning music while growing up (3rd world countries didn't have things like school bands or things like that 25 years ago) I apppreciate the fact that I can record a piano melody or a multitrack sketch of what I hear in my mind and send that recording to Suno so that it helps me recreate what had been already created by my brain but I didn't have the skills and knowdledge to transfer it over to the real world, and you can do that as many times as you want in case the AI didn't come up with something similar to what you had in mind. That in itself is a miracle to me, the next step would be having a hardware piece that could "hear" what you hear in your mind and replicate that using Vsti's, I'd love to see that because sometimes you have an idea and it goes away by the time you find the right sounds/presets/kontakt libraries that sounded like the ones in your brain.
These 3 major record labels are sueing these 2 A.I. companies to protect THEIR own greedy, fat bellies! They are taking the copyright argument too far. Nobody is reproducing their work without their permission. Training something to sound like something is NOT an infringement of copyright. They would need to change the law. Furthermore; these 3 major labels are not sueing on behalf of artists. For 17 years Universal, Warner and Sony have colluded and conspired with Spotify to con and SCAM artists out of BILLIONS of dollars! They did this by conspiring with Spotify to pay artists fractions of cents per stream. As long as these A.I. companies put some kind of warning or notice on the music saying "This Was Made Using An A.I. Artist", so that people are not misled, then I see nothing wrong here. The A.I. services must also be priced reasonably, at a point that most can afford. The REAL problem here is that Universal, Warner and Sony are investing in THEIR OWN A.I. tools and so they are trying DECAPITATE THEIR COMPETITION VIA THE COURTS!!!!
But he wasn’t spitting out pieces of the actual master recordings. The recording itself also has a copyright, and a human would usually have to clear samples of those recordings with the owners of the masters.
@@craigington73 it’s not about the result, it’s about how they got there. They didn’t license copyrighted music for commercial use when they scraped all the audio to train the algorithm. The law says if you use their product for any commercial purpose, regardless of the result without licensing it, that’s violation of copyright. The only exception is parody because it’s a first amendment issue.
The truth is that we could sue all the time because all of us are trained to learn in a certain way. And I think the AI breaks the barrier between the people that have the money and can easily buy all that they want and even the talent. The AI will help everybody to have a chance to grow and have a better life. These big Major Labels doesn't care of people, they just care of their money. They don't care about the artists; they just use them to win money, and now it is time for the artists to find their own way and be paid for what they create. What would happen if these Major Labels had created Suno and Udio and gotten the money for the abonnement we pay? I'm sure they wouldn't sue themselves. 😉
@@DigitalAI_Francky This is where you are wrong. The AI will not "help everybody" have a chance to grow a better life. The AI may help you make your dream album, but at the cost of its complete irrelevance. You think knowing how to prompt an AI will make you successful? That prompting is eventually going to run on autopilot, and produce an endless sea of content without a single word of human input.
Wasn't long ago people who saying that if you're drawing midi notes into a DAW, you're not a real musician. UDIO and Suno probably aren't even doing as much training as you might think. People, not computers are likely training the AI with likes and dislikes long before it was released to the public. Why does the music sound so real? Probably a web based VST where the AI loads the plugins based on the genre. Every genre has it's own basic formula...this has to be far easier to do than Midjourny or Stable diffusion. As far as creativity...prompts gets boring, you'll want to have more controls, so later you'll find meta tags will be used to control the output more and more, so as long as you're writing you're own lyrics, I wouldn't say you're not being creative. People will use these tools to help them out of creative blocks, they will bring it into their DAW's , and remix to their liking. If they don't then they'll sound like everyone else. SUNO nor UDIO isn't a threat to musicians. It's a new tool, but it is a threat to the Monopoly of the big three labels, no doubt there. Does not mean Udio or Suno are doing anything wrong though. Also those samples that seemed similar to other peoples works. Man there is no way those would even compete with the real thing. I'd rather pay to hear an original track than some dollar store sounding knock off. Even of those can be made in the system, nobody is gonna release it and if they did nobody would buy it. One last thing...if they are indeed analyzing other peoples works, it could just as easily be to prevent output that actually would be copywrite infringement.
A year ago the best song on Suno, was a complete A.I. joint about a Christmas turkey. It's only when Songwriters/ Arrangers and Producers, (the same talent that creates the goods now) started using the model, the interest, big money investments, lawsuits, and microsoft, rolled in. This thing been out a while. Nobody's interested in that stuff "casuals" put out in any field of endeavor. So all this talk about everybody making hits and blah blah, replacing blah blah, is pointless. Real musicians never made any money on studio work. They'll still make it like they always did--in concerts. Singers --the same thing. The Rihana's are still going to be getting theirs. Studio singers may get even more work because eventually even A, I. users, if successful are going to want a piece of that concert pie and instead of Rihanna, they'll get the studio singer to front their thing. Only thing that is going to change is the people are going to be getting some amazing music again
The irony is that EVERY artist uses this input-output method. We all listen to 100s if not thousands of different artists and genres that inspire us to create our own original musical pieces. This is not that different.
I agree totally. Musicians and artists are always inspired by other artists. 😂 and this is a good thing. You should be proud that someone use your musics to get inspiration.
exactly, all AI does is go many steps ahead and come up with a whole instrumental instead of going step by step as a human would (drums first, melody next, etc. It's simply like having a co/ghost-writter/producer by your side, helping you to develop your idea (I use it exclusively for the upload audio feature), and most if not all of the biggest names have had a whole team of producers and writters or a couple of them at some point.
That's not irony. The human brain learning from creative works, has been part of the social contract ever since the first art and music was made. A computer is not a human brain, and is not subject to the same limitations as the human brain that the original creators expected when first publishing.
Suno & Udio are making it possible for anyone to make music but unless the person is putting in the work needed, the generated songs are going to suck. Everyone knows Suno & Udio are trained on copyrighted works, but are they simply recreating them or making new songs. The answer is that Suno & Udio are creating new songs based on their knowledge. Composers fear that AI-created songs will replace them so they want to slow the adoption of AI music to make money before they are replaced. The record labels want money & power, this hasn't ever changed.
I’ve done some research about trained with copyright works. Finally, even the artists that are inspired by other artists train themselves with copyright works, just hearing it. The fact that they hear a music, they train themselves to learn how to reproduce a certain style. It doesn’t mean that it is a violation of the copyright as long as they don’t make the same exact music.
Yeah but this works like a printer. You could spill out like 200 Beetle sounding songs in minutes compared to one artist coming up with a few songs that have similarities. This crap will be way to powerful in a few years of it doesn't get put in it's place. One this is you can't copywrite this material, however, how do we know if someone is being honest that they didnt record a new song the they made with a prompt (unless all prompt data is public)? @fdupas
@@nwilt7114 sorry to break it for you but "honesty" doesn't exist in the music industry, just listen to the most famous singers live, most of them suck and lack either talent or skills, they are fabricated in the studio with melodyne, they have a team of people who are making their songs and many of those people are copying other songs to some extent or using the same sample packs everyone else uses, same for producers like dimitri vegas and those clowns, heck even those you would think were honest like armin van buuren are not. No one had this tech 10 years ago but you can bet that those fakes from big labels would be using it without telling to anyone if they had the opportunity.
Most of them clips you shown was from Udio... I use SunoAI and write my own lyrics, I love using it as I can hear my songs.. There is no way an Artist or a big record company is going to use my lyrics, I hope Suno wins🥰
It doesn't ingest. I hate the major record labels, they bring nothing but crap to music these days. Hence why I'll use udio or suno to make the songs I used to enjoy people don't make that sort of music anymore and the record industry is only interested in rinse and repeat.
And it isn’t so easy to create a good music even with AI. Finally. What is important isn’t how you create the music but if the result is good or bad. 😊
People can imagine a future where corporations own AI music platforms. They might continue to allow people to create and earn royalties, but through specific terms, they could essentially own all copyright information for the music tracks generated and add them to their own copyright libraries. Over time, people will find that it is impossible to create any musical phrases in AI music that do not infringe on copyrights, as the corporations monopolize every possible combination of notes.
@@holykylin I guess you don’t know that the US copyright office deems any AI output as uncopyrightable and therefore instantly in the “public domain” as soon as it spits out a result. In order to copyright, it’s got to be created by a human. There are case precedents where people have wanted to copyright art made by animals and machines, and they were declared uncopyrightable because a human has to do it. Copyright is for humans only. Educate yourself about this.
@@batautomat Indeed, I am not only aware but also understand that the United States is currently the only country that outright does not recognize personal ownership rights for AI-generated content. Other countries are in a state of ambiguity, or conditionally recognize such rights. However, this does not conflict with what I said. Firstly, the use of generated content isn't just confined to Americans nor is it solely distributed within the U.S. Secondly, while copyright laws related to AI may evolve with the times, private rights are almost impossible to alter. This means if copyright groups legally gain control of AI platforms, it would be nearly impossible to reclaim those rights. Perhaps antitrust laws could intervene? Even so, large corporations often have methods to circumvent these rules.
@@holykylin imagine all you want- it doesn’t mean it will turn out that way. You are using the same method that AI tech bros use, suggesting a future outcome to entice investors but maintaining plausible deniability if they end up being wrong.
I don't think they need to get permissions from copyright holders. Learning is never copyright infringement and transient copies are part of the internet, whenever we listen or watch a piece of media, the browser caches it on our computer, so making a transient copy to read something is generally not copyright infringement. However if the output is too close to existing copyrighted works, that's obviously a different story.
So is your position: "it's OK as long as they don't get caught outputting something too close to copyrighted work"? (Would you grant similar permission to someone taking things from a store without paying for it, as long as they only used it in certain ways?)
@@daviddas No because if you take something out of a store, it's gone from the store. You're effectively remove something. But I'd grant a similar permission to everything and everyone to closely examine an object in the store to determine how it was made and then use that knowledge to create a smiliar object in your garage.
@@NorthgateLP fair enough about a physical object removed from a store. But in the digital realm, there's always a transaction when a digital good is consumed. An iTunes music store sale would generate a 99-cent transaction. A Spotify play counts as a micro-payment. A UA-cam view counts as a view (and can be monetized). Because the digital good is under copyright, there's still a transaction that takes place. It's not just a glance at a product on a shelf.
@@daviddas Right, but if the AI companies pay that transaction fee to train their AI then I don't see why it should be prevented from analysing these songs and compose new music based on what it learned. One important thing that I see a lot of people misunderstand is that when an AI composes something new it has absolutely no access to it's training data and does not contain any copyrighted information in it's model. It creates things solely by what it learned through fancy math. Now if the AI companies have aquired their training data via illegal means that would be an entirely different story, but then the copyright infringement doesn't really have anything to do with AI.
If someone uses AI to mimic your voice and commit an illegal act, you will deny that it is not your voice, even if it is very similar. But if the sound is used to make money, You say it's your voice. And that's exactly what happened.
If AI even further decomposes the copyrighted material in their LLM to even smaller bits like notes would that alter things? You can't copyright notes, just like you can't copyright letters or words. It's a blurry line and I'm pretty sure the big music labels aren't suing because they care so much for the artists...
If they want to ban Suno or Udio, they must also ban all the text to image apps out there, and there are hundreds of them. All of them operate on the same principle, they look at images or songs online and compose new material. I vote for AI to stay, it is the future ♥
You're right, same principle. I don't know anyone who's advocating it to be banned; the only question is compensating the preexisting work that is used to make it what it is.
They should ban themselves to not let the people live their lives how they want. And even the artists are trained with the same technics with active hearing and music analysis, learning harmonies. But at the end, it isn’t sufficient to create a good music, the people need to like it too. If I like a music very much, I will buy it, and if I don’t like, I simply won’t buy it. 😉
if a song has like 5 or 7 consecutive chord progressions that are exactly the same then maybe. But just because a song has parts that sound similar, or overall sound reminds you of anothers does not constitute infringement... if a heavy metal artist does a vocal growl is that infringement?
Sorry, but there within their rights It is transformative they are not selling The music that they train on if I want to look at a painting that is copyrighted and make my own painting using the copyrighted painting as a reference that is transformative Sorry, Sony, but you’re wasting your time
I'm concerned that it might still fall under "sampling" in some cases. I hope they will only have to filter out songs which are very similar to existing ones, or pay to copyright holders a little bit if people listen to these songs on their platform.
@@NoidoDev it doesn’t it is completely transformed, if I listen to the radio all day, and I eventually sound like someone singing on the radio that doesn’t make me a thief , and just because technology is used to do this, doesn’t make it any less transformative sorry
Cant wait till they figure out how to train these models on less data. Aka only public commons, or self generated and rated dated. The sooner the ai midels can sidestep the copyright/ip paradigm the better
A problem I see is that it is at the core no different than what people do. People consume music, learn from it taking in elements from different genres and then output music that is often heavily influenced by it. Our brains are updating each time we hear a song. I Is that copying? My guess is that copyright is not going to successfully cover the training because it is technically not copying and we'll need new laws. It will be really interesting to see how this plays out.
AI is a great leveler. Studios, big labels and big artists can't tolerate an ordinary guy with interest in music compete with them. Also every human artist has "ingested" some music which influences what they create. So it's not fair to blame suno or udio. Lets face it. There is no such thing as pirely original music.
To be clear, it's the not the "ordinary guy" they're competing with. It's the AI. An ordinary guy only comes up with a prompt, which is not creating music at all.
This is the process that SUNO and UDIO uses, they train a model, by adjusting a neural net to replicate a close enough version of content based on the prompt given, but the way the neural net stores information, its impossible to ask it to give you precise songs, as it conceptualizes the the style of the music down to the similarities in the musicians voices and composition process. But to actually make the precise songs is like pulling teeth. If you don't believe me, just have Gemini or ChatGPT or read it from Wikipedia how a Neural Net stores information ,its like a fuzzy logic version of nested if-then conditions. Sending input in one end and gaining the results of one complex formula. And there isn't a person on the face of the planet that can identify what any part of it really does, its so complex. And its likely what the Lawyers of Suno and UDIO will provide the music industry lawyers when it asks for an explanation of where the content is stored.
Utilize this technology as a valuable tool to optimize your workflow and improve your output. Also, it is crucial for you to delve deeper into the workings of these LLMs. The level of sophistication behind their operation is truly awe-inspiring once you grasp the intricacies of these systems. They possess a remarkable ability to emulate the human brain, enabling them to learn from information and generate novel concepts. These companies that are filing lawsuits simply want control, whether due to a lack of comprehension regarding how these systems operate or a desire to profit from them, or perhaps both.
I look forward to how this plays out. I am under the impression that the way the companies trained their data is fine. I have a band, but I also write a lot of music. I think the ones who are worried about it are those who are composers, record labels, fill in artists ect. I am a singer who has a bad voice, and in my band I use effects and things to make it sound okay, I have used auto tune for years, and also beat randomisers, and other tools to get ideas. My fav tool is the band in a box pedal, but of course I don't get the beats from the pedal, once I have a functioning idea, I record it, then do the drums, bass, vocals around the guitar. With ai music generation, I have been able to get the syllables of my written lyrics, and prompt an ai gen with what I want. Then I work on that. I was also in a band with a lead singer who sounded like Shirley Manson, we didn't do garbage covers, but some people thought we did.
Well the reason why the lawsuit shouldn't really frighten anyone is the same reason why other platforms who do similar things shouldn't be worried either (or even be fearful of a similar lawsuit). For Example; Twitch implemented "Famous" voices as TTS voices for donations to streamers, trained on song artists voices, and mostly all (if not all) are trained on song data - The Weeknds voice is completely trained on one song; if you wrote down the lyrics for the song in your TTS, the song would essentially be played. Although the data is trained on these copywritten voices, anyone can get paid with the addition to these voices, adding formative content to their streams. Where is the line in this regarding copyright , and how does it relate to infringement as a whole for AI training data? well that's the whole point of the example, these things still need to be defined, and lines need to be drawn before estimations of crossing those lines can even be made. In the same way these companies are mad that these AI generations are creating things using previously existing art/ "their art" - the logic should fall under all matters of this magnitude, including but obviously not limited to that example. The problem with that, is that the very foundation of the lawsuit (This is data trained using our licensed music) Needs to first have a category to even fall under the guidelines of what is considered "wrong" of them to do, *first*, before they are even considered for doing "said wrong thing" within the confines of those guidelines. The courts need to define what they consider within the fair use-age regarding training data for AI as a whole. And THAT is a process that I don't think these companies realize just how long it is. Even if the AI companies are in the wrong, they first need to be defined within the confines of what type of copyright infringement the Music industries are claiming first. They might not even get that far. And this will take at the very least a decade to even draw the line of distinction. All in all, I believe this won't get very far. By the time they draw the lines on AI training data to even start this fully, the lines will be even further blurred.
It's tempting to view these lawsuits as a David and Goliath battle but when an AI model creates songs that don't just sound like the Beatles but sound like Paul McCartney singing on them then I can see why the major record labels may have a point.
I think what labels really want out of this is to protect their sampling businesses. If you allow the theory put forth by suno and udio to proceed without challenge, then transformative use, not just ai, but all of it, needs to be reevaluated, and they lose massive amounts of royalties they're making today. This might seem like it's about ai, but it's really about understanding the rules of the road, going forward, at least for the next 30 years or so. I'm glad they've opted to sue the corporations and not individuals though. Given the RIAA's history with this kind of thing, I'm very happy that they're going to a place where this can actually be resolved. No matter what happens, or who wins, ultimately, a decision here, any decision, would benefit everybody
@@daviddas there are also very good musics done with Udio. I'm also using it and I love it. I don't understand why they always complain about the fact that everything is trained with something. Even people are trained with something when we study, so what's the problem to use AI ? For me, that can be very helpful in everything.
If you can hear the song in your head and recite/play it, your brain made a copy of it. Treating AI differently when it comes to the process of learning through experience is (at the core) a human bias that creates the precedent that learning and performance is illegal.
Great Video David! We've created our youtube music channel few months back thanks to Suno and Udio. These AI platforms gave us a chance to share our creativity, even on Spotify. Something we couldn't possibly afford doing before.
I usually am not a fan of Gen AI, HOWEVER, the music industry is a plaque anyway. Michael J. and Prince already mentioned that. In this case, I actually would band with gen AI platforms, just to annoy those big wigs.
Atleast you music guys have these corporations fighting on your side, in the visual art space shit is ruined beyond repair. Mid journey has completely disrupted the entire industry.
Thanks for this video, David. I appreciate your dedication to exposing the nuance in a situation that from the beginning was intensely polarising, and remains so today. Please do continue talking about AI and its relationship to music so that we can get your take on any further developments in this space.
May be a hot-take but I think ai models should be allowed to use any material they find on the internet as long as that material isn't directly included in the product result. This is what we humans do, we take inspiration from any and everything we can. I'm allowed to look up a picture of the mona lisa, study it, and lit that knowledge influence my artwork. Hell, I'm even allowed to paint as near of a replica as I can manage. AI is designed to learn in a very similar way to how we are. It will outperform us in every way eventually, slowing it down doesn't really change anything. A bunch of rich people are looking for ways to cash out on some abstract controversies, nothing new.
If this lawsuit is successful, it means all ai tools will have to be pretty much rebooted as they all borrow and learn from all available material. Suno has already hit back at this pretty hard. I'm also wondering what ai tools Sony, Universal and Warner are using. If they use any at all, it would be good to point that out.
If "kinda sounds like and reminds me of" becomes a copyright violation.... kiss the act of making music outside of the permission of the big record labels goodbye.
God forbid something that a major company doesn’t own exists. The lawsuit won’t help any musicians. This will only put more money in the pockets of the greedy people that ruined music decades ago. I’m mean seriously when is the last time you turned on a radio.
All music created is based on previous works essentially. As a musician I grew up studying and learning music and those influences absolutely are present in the music I create. The only issue here is Industry Rule #4080, Record Companies are shady. Embrace the tech. The best music will have human input. Prompts are no different than a producer telling session musicians what to play.
If you say that "users of Suno and Udio aren't making musing, they're entering prompts", then you could also argue that "the person who plays the piano isn't making music, he's pressing keys". I understand that AI is heavily disruptive to the industry of creators and distributors, and that society might want to think of a way to protect those who could be harmed severely by that. (I would support that idea a lot.) But the current arguments to sue software companies for their technological innovation seem quite ridiculous. You can easily copy a painting by taking a picture of it, but until now, it was always the responsibility of the photographer... and when trying to sell that picture, he should be held accountable if that would be a copyright infringement - nobody would think of suing Canon or Nikon or another camera manufacturer. Why is it so hard to treat the current audio creations in the same way? When a song created by one of these tools is considered a copyright infringement by a court, then the person who tries to make money from it should be the one to blame.
They ban others from using it, but they will use it. They got big appetite to feed themselves. We all know that. I’m using Suno and Udio, they are good. Traditional music producers need to grow and move along. This is the next phase in music creation. Hate it all you want but this is the next direction, or be left out. It is developing and learning, and it is growing and getting better by the day. It’s a hit and miss to get what you really want but you do get nice ones. It can be used to find ideas and re-record nice parts, and different parts together to get a full new nice song.
I hear your arguments as a composer and I respect them. As someone who found Sumo this week, I;''ll add this. I'm a wannabe fiction writer so I respect the position of 'Creatives. I've also fancied writing lyrics but I have no composing ability... at all.. I'm not just using prompts in suno, I'm writing my own lyrics, or changing the ones it creates so its 75-100% my words... Then generating the tune and performance. If i take my work anywhere outside the 'Suno community' and my friends, I'd like to think I could find a composer who could write new music or arrange what I've created... But these are early days of my relationship with Suno, so who know?
Well-known songs and artists are imitated by stock music composers/producers who tag their tracks with genre, mood, style, BPM, and sound-alike keywords/names. If a generative model was trained on that data, it could output music that sounds very similar to the original well-known work without ever having directly ingested it.
Thank you David...your video was AWESOME and you explained it very well. I teach news related English in Korea, and this was a story we did in class yesterday. I tried my best to explain it but I shared your video with my students as your explanations are perfect and clear with some great examples! Fantastic video! Cheers! 👍👍👍
If we look into history we find that many of the real masters behind the real artistic abilities were many times caught into contracts and were never really paid what they were worth.. In addition some of songs we hear and know today were stolen and made popular. So my view is it was never about protecting the artists, it was always about protecting investment, and margin and generating turnover. Many of the artists have aired those facts over the years and I can give specifics... Look at the classic composers of generations past, be it music or art.... A lot of that music and art was never paid and the artists died poor or from sickness but their works sell for millions today. In addition any type of music has rules built into it and every dictionary is limited . Music is not unlimited as many may think because the life experiences within a lifetime are recursive in each new generation....These are the subject matter for any good music... So the record labels needed to see this change. What they have done to other artists are now happening to them in exactly the same type of odd way. They are getting cut out of the picture and I find it amusing that they are complaining about that.....The solution however is to find a better way for artists to share their music without the bigshot labels.Take away that waterballoon and probably the music will be affordable and the artists would earn what they deserve, so that the revenue goes to the artists and not to the labels..... We use AI today to find answers we dont have... as a helpful tool... Did we not think that one day someone would come along with ideas to tip the balance back into the favor of common folk ? Supporting and enhancing the ideas of artists that would have remained unknown to the big labels. I also however draw the line that the real artists should get a piece of the AI pie if it learns / feeds on a library of already made music--- As for the labels you can make anyone into a star just by having the right tools, however taking music that we all love, live and work by, it being more a heritage, and then holding that ransome by moving it into a money making business, also means that those who cannot pay dont get to hear it..... Or artists that dont pay the big labels will never make stardom... I´ll leave with another thought, "How long will it take the AI companies before they decide that their tool makes them part owners of the art / music in the future.... Maybe we should find the person that made the paintbrushes that M.Angelo used to paint with and pay him billions for part of the sucess story.... Maybe someone can create a song about reoccuring life experiences and put a patient on that...... as if no one else would ever have that experience or come to any of the same conclusions..... ;)
I think it’s funny lol. A.I don’t have any idea of what is actually is excepted. It follows commands and what ever you feed data wise into it. Udio can be changed and made where the voices don’t have to imitate a existing artist. Maybe parameters would be useful
Incredible that this techies even ADMIT that they completely ignore EVERY SINGLE LAW that (at least marginally) protects working artists like myself. And what is most depressing? I have plenty of friends/colleagues/etc who use AI art and music frequently, for work, for pleasure, f.e. kids birthday and so on = it is a self-destructive cycle!
Record companies are finally reaping what they have sown. For years, they have exploited artists, profiting immensely from their talent while giving little in return. However, the advent of new technology heralds a transformative shift that could drive these companies into bankruptcy. Their arguments against this innovation are utterly unfounded. Artists have always learned from one another, drawing inspiration from chord progressions, rhythms, and harmonies. Now, AI enhances this creative process by providing us with prompts that lead to original work. There is no fundamental difference between the ubiquitous use of the same four chord progressions in countless songs and the use of AI in music creation. I find it amusing that the record industry, after years of exploitation, is finally facing the consequences of its actions.
How is this any different than image ai systems trained on copyright images? Or text systems trained off scraping the web and other copyright material???
You're right on the money. There is a mistake in saying that these aplications "listen" to music or they are "trained" by music. So later on they imitate those recordings parameters. Becase they're not living consciousness. They don't listen to music, they incorporate those recordings into their databases and use them in order to operate. That is commercial use of a copyrighted work without a license. Copyright infringement 101. We don't need more laws, we need to enforce the copyright law that is already In place and ban these applications for good. All those companies would be out of business if they are obligated to pay just 1% of the penalties of the violations they've already done. And there is no safe way to let them be because the ripped off has already been done on a massive scale. And I agree with you that the major labels don't want to eliminate those so creators can make a decent living on our creations. Otherwise they would have never give away their masters to the streaming platforms. They were the ones sponsoring the streaming monster while telling us that Spotify is the bad guy. (I'm a songwriter)
That's not how they work. Effectively they listen to the songs. The songs are input into the machines and the machines update their parameters to predict what's going to come next in the song. And they are trained on a large volume of songs and end up learning parameters that can generate a large variety of music. It is "training" and not really so different from what people do. Of course though like with expert musicians, if you tell it to reproduce a certain style it will be able to. That is the violation of copyright and would be a violation with people as well. Considering the training of the parameters a violation would have numerous potential consequences that are not clear. Don't get me wrong though. I am not arguing for it or against it.
@@ssssssstssssssss They don't have ears, so no, they don't listen. They record, copy or scan audio files and incorporate all that into their database. the softwares are "Using" copyright material to operate without a license. that is copyright infringement. The trick is to fool the people making them believe they are living entities so they do the same as we humans do. No matter what wording you use the outcome is the same. Feeding their databases with all of our works in order to operate. Using copyrighted works commercially without a license. And all of this is agravated because it has been designed to put the creators out of business to push political agendas. #1-The Digital ID, #2The Universal Basic Income. This is criminal against humanity. Besides if you want to be more credible change your profile, make it look a bit like there is someone behind not a bot or a sponsored troll.
@@NoidoDev They operate using those copyrighted works so they can spit those approximations you mentioned. No matter how they use those copyrighted works, they're using them without a license. Copyright infringement 101. But even those approximations you mention are so poorly done that you can easy recognize the songs they're copying and in some cases they even copy exactly the same lyrics of famous songs exposing themselves as plagiarism. If the law is enforced they will be out of business inmediatly. But there are political agendas behind these Ai apps.
@@DanyTomaslifehave you ever used any of these ai programs? I cant tell anything about udio but i have experience with suno. I spend hours just to make a song exactly how i want to sound like, and no, it never sounds like any music i listened to in the past from real humans. Congrats for taking away the chance from poor people to have some fun and create their own music for themselves. I know it has to be regulated somehow, but the sad thing is that the big music companies want their profit and position, or if they really cared about the artists we wouldn't have such a horrible music industry. That's why I started to make my own music for myself. I am not going to support something I am not satisfied with, aka the shit I can hear on the radio again and again. I am gonna support what I like, and I am going to listen to the music I like. Suno made it possible to me, period.
Read the terms and conditions of Spotify or Apple Music, or the back of any CD and it says "Non-profit use only." Training AI on copyrighted music in order to sell the music it creates is for-profit. Ergo, all of Udio's output is in breach of copyright. They need to be sued out of existence!
Isn't this literally how Spotify was created? The creators needed music so they took it without permission and got forgiveness AFTER the fact, and now the major music labels are part owners of the platform. So don't be surprised when these labels end up partial owners of these AI platforms. Suddenly they have their "cut" and aren't mad any more. Meanwhile artists will get next to nothing.
The fact that I can show you or anyone an entirely AI generated song without them having any idea it was not human made disproves anything you can say about the difference between AI and humans. We are literally there. AI can be entirely convincing as an actual person.
I'm with you 100%. 'Fair use" NEVER meant 'for fair profit'. Educational and news commentary, and only in segmented clips is what Fair use is meant for. I hope these guys are sued to oblivion. If anyone can push a button and get a competent song, then music becomes worthless. AI is strictly for untalented people. It's bad enough with spotify giving anyone access to all the music in the world for eleven bucks a month and now we artists have to deal with this?
I got a few minutes in to this and up pops an advert for AI generated basslines! It asserted that back in the day when people started using laptops and software people using big consoles and tapes said that was cheating. I don't remember that. Computer stuff was actually alot more expensive and the evolution was analogue, then these huge digital systems for hard disk recording that only major artists could afford. Everyone else still used consoles and tape until the costs of digital came down. No one said it was cheating.
The thing is any human can copy a song and it is humans that are creating these similar songs.. The law should not be based on how the AI learns it should be based on how the end user uses it. There needs to be more safeguards to prevent direct replication of songs whether it be instrumentation or vocal melody or likeness of the singers voice while still retaining the knowledge of previous work to inspire future music. Exactly as humans have done for centuries.
I have plugins in my DAW which handle vocals I’ve written and another ones to suggest bass and drum lines after analyzing singer’s voice. After that I can sidechain other plugins to generate guitars and strings, etc… Yet this is considered to be my own composition. Having tried Suno, it prodeses soulless radio formats without real knowledge of song’s meaning. Sort of nice to try if the genre you thought of works, but it’s still less stressful to have real musicians singing and playing them as they are intended.
As a musician, I spent my life never listening to any music - only writing and performing my own. Otherwise I might have been influenced by that music and broken copyrights. Maybe that contributed to my complete lack of success as a musician.
Spot on. It's just a fancy 2024 way to rip off artists. "instead of entertaining a good faith discussion, they've reverted to their old lawyer-led playbook." Yeah, well Bubba, that's what labels do, they protect their intellectual property.
The funny part is Udio and Suno said they "care about the artists, that's why we ban artist names in our prompts", when they are doing it to hide the fact they've trained it on copyrighted material. Practically every song on Spotify is copyrighted. Regarding your final comment on everyone getting a share of the AI music, that only applies to artists on those labels I'm sure. The only good outcome for indies is these companies will be stopped. Anything else will still mean a cheapening of artists' work, ie the prosecution's argument will most likely still apply to any artist not already on a label.
Great video, as a music and video creator I was wondering when, and hoping for, the day these AI companies get taken to task for nothing less than outright theft. As you say, they are not making 'musicians' either, it's just a load of talentless people typing prompts - it's only going to homogenise music not be creative. I wonder when the big image companies (say Getty) will also take action. What should happen imho is the AI companies should only work from a licenced pool of music/images - maybe artists can freely upload into those pools. When ever any of the music or images in that licenced pool is used (and they can easily track that) for any AI then the copyright owners/performers etc get credit and a financial proportion of the AI company profit. This will mean AI won't be free, but it shouldn't be if it's just is parasite feeding off others people's creative endeavours. We had the same sort of thing when mp3s were being generated and shared for free, it seemed impossible to control at first but once the law cases kicked it it got changed.
Udio isn’t going to take your composing jobs, Sony, UMG, and the rest will. Udio users won’t replace anyone lol. They busy making music about farts or cats.
The second artists start licensing their style and works to be used, every song will be just another copy. Everyone will be making their favorite artist's work. Originality... gone. The protection so to speak that Suno and Udio have in place that blocks the people who, as you put it, prompt, is there to keep people from doing that and it keeps the work original. When some label leases out the rights to like a Billie Eilish song, every song is You Should See Me In A Crown.
I think the vast majority of these would be classified as inspiration. If you have specific examples of "stole"n music, i.e. theft, you should post specifics. Every musician who has ever lived has been inspired by artists who came before them, whether they be from their own tradition or another.
What is the difference between an AI model ingesting any recorded music, data on music theory and an artist spending days, weeks or months listening to the same music and reading endless articles, books etc on music theory and sheet music a being influenced by it to write, record and perform something similar? My view why these record labels are only going after Udio and Suno rather than other players in the AI music generation market (even so called ethical players) is because there is a difference between them. The likes of Soundraw, Soundful and others only produce audio output and backing tracks at best because you don’t have any real control over the output, but Udio and Suno standout because they can produce the audio in a way that is more creative and a kin to actual songwriting by using meta tags for thing like intros, instrument solos, voicing styles. Add to that the addition of vocal output then it is a danger to their business, some would say to artists as well. But to me an amature creative with some musical skill and ability to write lyrics they provide a tool to produce something that not so long ago I could only hope to achieve. I also find it strange they are not suing the AI voice changers in the same way, as these provide an equal threat to them and their artists because they allow recorded vocals to be swapped out for somebody else who didn’t actually record the vocals, but maybe the reason for this is that they are actively investing in these voice changing AI models so that they can exploit them and their artists by being able to create new Beatles songs or have Billie Eilish cover a song she doesn’t want too.
I think it's great that AI creates music that "sounds a bit like someone else's music", but how do humans use this song? The moral standards are still firmly in the hands of humans. If the similarity reaches a certain standard, it will be labeled, just like the rating of movies.
I understand why song writers and musicians are bothered by this but I don't think there is much that can be done about it. Automation is going to continue to replace the need for humans to do certain jobs. I do not think A.I. generated music will ever fully replace human creators because we all like genuine talent but it will compete.
@ 4:48 Mikey Shulman saying "Suno's mission is to make it possible for everyone to make music. It’s like he saying “ My Wife is for everyone anyone can butt f**k her.” It’s just like that 😂
see my problem is that i have specific type of music that i like and its basically impossible to find good artists making those so im going to still make those myself and share them
This lawsuit looks like to me a way to shutdown the competition that the big record labels can’t no longer compete. I completely agree if the content is copied, the creator has to be cited, compensated, etc. I use Suno, but I write my own lyrics. The songs crated by Suno with my own lyrics it was never copyrighted. There we go! You mentioned about AI replacing composers like you, maybe you should try to incorporate AI to your advantage and keep up with AI. AI is here to stay.
Thing is they’re not doing this because they care about the artists. They’re just mad they didn’t do it first. There are already major labels using signed artists to train vocal models. They’d be happy to shovel generated content into our ears is THEY make it and profit from it. It’s not about ethics, or artist rights, at all.
It's funny when all these music people are fighting FOR the big music labels to control the music industry even more. They want "kinda sounds like and reminds me of" to be a copyright violation going forward..... it's crazy.
If the major labels put out a product like suno or udio... I would be a vip for how often I use it, and what I do with it. Just like I am at the other two. I think it's a great idea, and if paying them is what it takes, fine. Paying them is no problem. That said, I don't think they could do the job of training that Suno did without help. The breath of it. It's more than music or tags. That's where most people are getting this wrong. Suno has things in the training data that make it a musical instrument... if you know how to do it. I believe that's because they scanned a century of sheet music as well... among other things. Udio has better fidelity, but theirs is a much simpler training than the Chirp models.
Napster all over again :))))
Record Labels had made a fuss about mp3s.. because it was free.. then the bought Napster, then came Spotify.. now everybody listen digital music.. nobody listens CD or Casette :)))
Generative music will prevail eventually... even now, some musicians and producers are using this tools.. b/c it gives creative ideas and precents writer's block.. and everybody with a simple idea can become an artist.. no ne to know notes and sheets and play any insturments and have a great voice :)
Agreed!
Fine shandy
But the major labels want to create their own AI music and profit from them. They’re not protecting artists. Spotify wants to create fictional artists for playlists and collect the profits.
Yea and it is no secret they have that one Swedish composer to write all the easy listening piano tracks. Look to find the names they serve as the composers of each individual track and good luck finding a trace of any one of them! These are fake names. They have made 150 million dollars from those tracks and paid him 6.7 million.
They're protecting their assets. Who wouldn't?
@@BrianLarney paying the actual composer that less :)and also faking names ...yeah, sure, protecting assets :)
Major record labels will lose and can’t stop AI from training on music copyrighted or otherwise.
@@RideShareRocks I’m glad you can predict the future- you need to market that. Face it, you don’t know who will win. Gen Z is already onto AI music and is rejecting it outright. They have good ears for what is fake and they hate fake. I work with a lot of of Gen Zers and once you lost them, good luck. They are the future of culture and the ones who are ultimately in control because they will refuse to consume it. It will become the equivalent of any product that won’t sell. There was no long term business plan in the Udio or Suno training. Outright rejection of the apparently easy to spot AI songs is something they didn’t factor in, but it could be the fatal blow to a grand experiment that no one asked for.
Yeah, I'm sure those major labels are doing it to help the artists. I'm sure.
😂 Right they are worried they won’t be able to pay back the Loan they gave the musicians. “Record contract” 🤣👎🏼
I think if used in the good way, the AI, especially udio, can help artists that they are in lack of inspiration. Then real artists can go touring and make money with that... AI can't go on tour and for sure the people using AI to create music at the moment are like me that I am so out of tune, that if I sing under the shower, the water will stop and refusing to work again 😂, but sometimes I have some nice ideas that I like to see how would be. And now AI help with that. Isn't that great?
If an artist is really good, he will make money with live concerts. 😊
It’ll be interesting to watch how this plays out.
People are also trained on listening to existing music and they also generate music the same way AI does. This is inevitable. These companies are in a state of urge because they are afraid to lose the monopoly they have in the music market. There is no way they can escape from what the future holds, though. They have to adjust, find new ways of earning money from the backs of the musicians or they will not exist in the future. How sad they are not willing to be a part of the future! They still want to direct the trends in the market and direct the people’s taste for music by their huge advantegous hands. But behold, this will change. This will change for the movies and so on too.
It’s very different. These platforms are trained on the actual master recordings, not just being ‘inspired by’ the melodies/lyrics (which would be fine). They are essentially sampling, and usually a human would have to clear that sample with whoever owns the masters.
@@bw2937 I think human mind is also trained on listening to the same masters. You can even not knowingly replicate an album’s or band’s sound when mixing and/or mastering. This discussion is no more than being arbitrary unless there is full scientific research covering human vs AI learning/ interpretation/putting into use with every parameter and aspect compared. Even this will be the first step into making laws accordingly. Generative AI requires a different legislation then the current one. Until then even these lawsuits will result in favor of them, there is no way escaping what is to come next.And AI does not use the masters as is.
I thought. Exactly the same about most of the artists are also trained by this way too in using active hearing and music analysis. This big companies like Sony are just afraid to lose a lot of money. They don’t really care about the artists. For them an artist is just a product as an other one that they just want to sell to win money. I think that AI is a big and good revolution for the civilization. And finally, it isn’t easy to create a good music even with AI. I think that if a person is able to create a wonderful power ballad or other great music or song with AI and that people like that song, then he also deserve to win some money for that and not necessarily these big companies.
@@DigitalAI_Francky I agree, Sony once gave me a strike and wanted to scare me for allegedly using their music. The funny thing was that it was a recording of a Mozart piece of music with my music club, although the sheet music arrangement was already well over 100 years old. Of course, I had proof that it was a copyright-free piece of music and the strike had to be withdrawn.
look at the advancement in sound sample libraries to get a sense of what's coming. Sitting in music theory class in 1977 my instructor (with a PhD) said, "one day these synthesizers will replace musicians'" At that time it was only analog synths. I thought his statement was ludicrous.Look where we're heading. It's really the old story of Frankenstein, isn't it?
I can already see myself making grandma's birthday song in the darkweb.
AI doesn’t need to legally obtain anything to learn from it and it never will.
The problem is that it does not really matter how the music was made. That is an old way of thinking. We are creating a new way of making music, and the old way stands in the way for progress.
I agree, what is important is not how you create the music, it is if the people like your music or not. If they like, they will buy it and if they don’t they won’t. It is simple. I can buy a music even if I know that I could hear it forever for free in UA-cam but I pay because I considere that the artist deserve to be paid for it.
The AI that UDIO and SUNO uses is based on Musicgen's related technology, which uses a multiband diffusion approach. Musicgen is a open source technology that can create music from prompts as well and it can use a melody from music you upload. I have two tunes where I took the melodies from "Hide and Seek" of Howard Jones and "Desert" by emilie simon (a french pop artist). Believe me, the technique is so well known among the AI world that its not going to matter a hell of beans if the music industry succeeds in stopping SUNO and UDIO, its just going to entice hackers and activists who want to eliminate the record companies cause of the evil behaviors they have exhibited in the past.
As a musician and writer, we must try not to be affected by music that we love in a way that it's to duplicative, and anyone that does write music knows it's hard not to do this. Why should A.I. algo's be any different? Now I'm going to go all rant...
These labels have a major fight on their hands. Frankly, they have had it too good for too long, turning good wholesome roots into a teenage "poppy-mill" with roomfuls of writers and pre-teen sex object singers that can't even write their own name. It's about time these moguls poked their heads out of the money pile to have a look around. Competition is scary huh? Boo!
If someone can create a beautiful music or song with the AI, I dont’t really see that it is a problem as long as many people love this music. If I like a music, I will buy it even if I know that I could hear it for free in UA-cam. I like to buy it in ITunes for example because the artist deserve to be paid. Now, I don’t know how much are artists paid from these platforms but at list I pay and I find it normal. The AI will never be in competition with performers artists in live concerts at least and this is not so bad. But it isn’t so easy to create a good music with the AI. Of course you can create very beautiful music too but it takes also a lot of time and try before to get a good result. Everybody want to talk about freedom, so it is also a freedom to be able to create music even just for ourselves too if we want. 😉
I had Udio generate music for me to go along with lyrics I provided with very little prompts otherwise. Out of nowhere, it lifted the main piano riff from Gary Jules' cover of "Mad World" note for note but in a different key. If it had merely shifted a few notes I may not have even realized where it got the sound from.
This result surprised me, as I had assumed that the reason other people were getting copyrighted infringement results was due to their prompting, telling the AI to "sound" like something or by providing very similar lyrics. Basically guiding it and/or limiting in ways where it couldn't help but copy something. But there was nothing about my lyrics that matched the original song in words, theme or structure, and it only used the riff when generating the song for me, the melody for my lyrics was entirely different. I gave it pretty wide latitude to create something entirely new and yet it directly copied a significant and defining portion of another song.
I've come to realize that generative AI is simply a very splintered plagiarizer. It takes tiny fragments of existing works and rearranges it or modifies it until it's largely unrecognizable. But if it lacks enough variety, the chunks come out bigger and CAN be recognized. It's just a very sophisticated imitation machine. It does not "create" anything, it simply mimics what it ingests.
This differs dramatically from what a human does with the content we ingest. We "feel" something and are inspired to create something new from the emotions the music generated, or we understand what the artist was going trying to do and build on that idea and take it somewhere new. AI has no understanding at all of what it generates, it merely identifies patterns and reorganizes them in a way it predicts will be successful. It doesn't even know if what it generated is any good, it only predicts that the odds are good that it might be.
I used to think "what's the big deal of AI learning from existing works, humans have to as well, AI just does it much faster." Now I understand that, no, generative AI is not really "learning" anything. You're just giving it more content and patterns to copy and/or imitate. Whereas a human runs the risk of plagiarizing if they're being careless or lazy, AI can ONLY plagiarize.
That's your opinion. Tech is going to evolve as it always does and make people's life more comfortable and give access to use certain things easier. AI draws instead of you, write novels. Am I bothered by it as a hobby novel writer and artist? No. It's all about self expression, the market is there you can choose what you want to see, read, listen. If you are good in what you are doing, with some brain you're gonna get the attention.
Udio is all over the place.. same prompt may end up creating dumb tunes.. or create realistic artistic songs even with real people's voices..
i guess Udio is more trained by actual copyrighted songs.. but it was rather a new release than Suno.. so it is not great than Suno... but it definitely copying real music (i m down with that :)))
@@HERKELMERKEL im not familiar with udio, but i spent hours and hours just to create a song in suno. Never had a feeling its a copy from existing songs.
@@Reealos I am Turkish and i accidentally created exact copy of at least 3-4 of Turkish singers via Udio and yes Suno is not creating a copy.. i am using since discord times.. and i have 100+ song that i have really liked and saved
Crazy Train - ozzy
Sweet dreams - Marylin Manson
Undead - Hollywood undead
These three things have literally almost nothing in common song wise... except a little micro-plagiarizing they do throughout the entire song. Now... Crazy train may have come out with it first, but do we really know if they're the first?
The vast majority of rock sounds like rock because people do similar things, micro-plagiarizing in a similar way to the AI generative models.
Humans are NOT creative, we are literally the sum of our experiences.
As a fellow musician, I understand your bias. However I think you're overlooking a dramatic difference between Suno and Udio. Udio showed up several months after Suno had completed training on several versions of their models, gradually and with intention homogenizing their output to NOT sound like any existing artist or song. I strongly suspect that their most recent model leans heavily upon licensed studio music. Meanwhile, Udio has not had the time to move beyond the sound of the music initially used for training. Udio is much more cavalier about the music they generate. I think that the music industry is going to have a very hard time proving (by generating songs) that Suno contains any specific song. I've tried hard to trick it. If you try inputting existing lyrics, they never sound like the original song, even with audio input, it doesn't happen. And, Suno makes a pretty strong effort to refuse copyrighted lyrics, too. It will be interesting. But, IMO, you're more likely to violate copyright with your DAW. This is relevant because the only way to even attempt to violate copyright, using Suno, is for the user to intentionally push the program to do so.
I've tried hard to trick it too, and haven't been able to, but I've had friends of mine who have. I have no particular comment on the differences between Suno and Udio as I've tried so many experiments I've lost track of what worked best on each platform.
@@daviddas you can add [chord] annotations to mimick a progression, but Suno probably still won't deliver anything that resembles the original song. In all honesty though, I've occasionally gotten random voices that sounded very familiar... Udio, on the other hand, will gladly imitate the Beatles, etc.
the labels already said that they are in the works of creating their own "models". they just don't want to fall behind again like the whole Napster thing. They don't care about artists.🤣
Ingesting works is what every artist/creator does. This is not copyright violation. Producing new works based on what was ingested over a lifetime is what every artist/creator does. This is not copyright infringement.
Forcing/coercing an artist/creator to produce a work that is a forgery or copy of an existing work is illegal. This is what the record companies have done, using exact lyrics and artists' names when using the AI tools. They are the criminals here, not the AI tools.
Spot on ! Humans are doing the exact same ! As a musician my self I can say all musicians start by copying what inspired them in the first place. We generate music this way. They just scared and greedy , they see the writing on the wall.
I've been writing music for decades... not ONE completely original idea LOL. Nothing comes from nowhere.
You're anthropomorphizing. Humans don't function the same way; they retain ideas, feelings, and their passions. The human brain acts as a big filter, whereas an AI's 'brain' merges information extensively - and therein lies the issue. With the right input, AI can produce something very close to the original, albeit with some loss. It's akin to recording songs from the radio onto cassette tapes with hiss and wobble when I was a kid. If I remix two songs from those tapes, is it my creation, or does it remain a remix of existing songs? This concept often eludes people who struggle to grasp how a remix can sound vastly different from its originals, or how a billion songs can contribute to just one remix.
@@nickmack7282 Humans are filtering AI outputs though.
@@High-Tech-Geek are you a federal Judge who is knowledgeable in copyright violation? I think not. It’s really in the hands of those whose job it is to interpret the law. And it could very well go in favor of the copyright holders. Scraping copyrighted music is theft. The law says you must license the music to use it commercially. They didn’t license it. You just don’t know how this will turn out, period.
In the ongoing copyright lawsuits against UDIO and SUNO, we are witnessing a possible future where major music corporations could dominate the AI music scene. If these companies were to claim in the future that since they own the copyrights to the music used to train these AIs, they should also control the entire AI model, then we could see a significant shift in who controls this technology.
If these corporations win, SUNO and UDIO will likely fall and be subject to buyouts, further concentrating power with these large corporations. They would then have exclusive rights to AI technologies capable of producing tracks that compete with top artists. This control could severely limit innovation, making it too expensive for new musicians and small businesses to access these tools. The result? A lack of diversity and creativity in music, with the entire AI music industry-or perhaps the entire music industry-falling under the control of a few large companies.
We must recognize that allowing a few companies to use copyright laws to monopolize advanced technologies threatens artistic diversity and is against the public interest. Music should be a global treasure. We call on everyone, including policymakers, to focus on this issue to ensure that technological advancements are shared fairly, benefiting all, not just a select few.
As a user of Suno A.I. that used to make music with my keyboard MIDI with FL studio over 20 years ago, it's incredibly fun and powerful tool to realize a vision you have for a song, if you use the custom setting you can write your own songtext and create a mix of genres you are looking for especially if you are very descriptive, sure expect a few hundred retries to get the exact result you want. But this is for sure the future of music creation.
As someone who didn't have the luxury of learning music while growing up (3rd world countries didn't have things like school bands or things like that 25 years ago) I apppreciate the fact that I can record a piano melody or a multitrack sketch of what I hear in my mind and send that recording to Suno so that it helps me recreate what had been already created by my brain but I didn't have the skills and knowdledge to transfer it over to the real world, and you can do that as many times as you want in case the AI didn't come up with something similar to what you had in mind.
That in itself is a miracle to me, the next step would be having a hardware piece that could "hear" what you hear in your mind and replicate that using Vsti's, I'd love to see that because sometimes you have an idea and it goes away by the time you find the right sounds/presets/kontakt libraries that sounded like the ones in your brain.
Thing is, music is not just "learned" like you learn carpentry or auto repair. It is about 90% talent, 10% learning.
GREAT job with this breakdown of the issues David! Just stumbled upon your channel this morning. Do you do Sync licensing as well?
Yes, a lot of the music I write is for sync.
@@daviddas Very cool. Keep up the great work on your channel 👍
These 3 major record labels are sueing these 2 A.I. companies to protect THEIR own greedy, fat bellies! They are taking the copyright argument too far. Nobody is reproducing their work without their permission. Training something to sound like something is NOT an infringement of copyright. They would need to change the law. Furthermore; these 3 major labels are not sueing on behalf of artists. For 17 years Universal, Warner and Sony have colluded and conspired with Spotify to con and SCAM artists out of BILLIONS of dollars! They did this by conspiring with Spotify to pay artists fractions of cents per stream. As long as these A.I. companies put some kind of warning or notice on the music saying "This Was Made Using An A.I. Artist", so that people are not misled, then I see nothing wrong here. The A.I. services must also be priced reasonably, at a point that most can afford. The REAL problem here is that Universal, Warner and Sony are investing in THEIR OWN A.I. tools and so they are trying DECAPITATE THEIR COMPETITION VIA THE COURTS!!!!
But it is an infringement of copyright when pieces of the actual master recordings are spat out. A human would need to clear a sample like that.
Elvis Presley was "trained" on black gospel music. Should his estate be sued for millions?
But he wasn’t spitting out pieces of the actual master recordings. The recording itself also has a copyright, and a human would usually have to clear samples of those recordings with the owners of the masters.
@@craigington73 it’s not about the result, it’s about how they got there. They didn’t license copyrighted music for commercial use when they scraped all the audio to train the algorithm. The law says if you use their product for any commercial purpose, regardless of the result without licensing it, that’s violation of copyright. The only exception is parody because it’s a first amendment issue.
The truth is that we could sue all the time because all of us are trained to learn in a certain way. And I think the AI breaks the barrier between the people that have the money and can easily buy all that they want and even the talent. The AI will help everybody to have a chance to grow and have a better life. These big Major Labels doesn't care of people, they just care of their money. They don't care about the artists; they just use them to win money, and now it is time for the artists to find their own way and be paid for what they create. What would happen if these Major Labels had created Suno and Udio and gotten the money for the abonnement we pay? I'm sure they wouldn't sue themselves. 😉
@@DigitalAI_Francky This is where you are wrong. The AI will not "help everybody" have a chance to grow a better life. The AI may help you make your dream album, but at the cost of its complete irrelevance. You think knowing how to prompt an AI will make you successful? That prompting is eventually going to run on autopilot, and produce an endless sea of content without a single word of human input.
Wasn't long ago people who saying that if you're drawing midi notes into a DAW, you're not a real musician. UDIO and Suno probably aren't even doing as much training as you might think. People, not computers are likely training the AI with likes and dislikes long before it was released to the public. Why does the music sound so real? Probably a web based VST where the AI loads the plugins based on the genre. Every genre has it's own basic formula...this has to be far easier to do than Midjourny or Stable diffusion. As far as creativity...prompts gets boring, you'll want to have more controls, so later you'll find meta tags will be used to control the output more and more, so as long as you're writing you're own lyrics, I wouldn't say you're not being creative. People will use these tools to help them out of creative blocks, they will bring it into their DAW's , and remix to their liking. If they don't then they'll sound like everyone else. SUNO nor UDIO isn't a threat to musicians. It's a new tool, but it is a threat to the Monopoly of the big three labels, no doubt there. Does not mean Udio or Suno are doing anything wrong though. Also those samples that seemed similar to other peoples works. Man there is no way those would even compete with the real thing. I'd rather pay to hear an original track than some dollar store sounding knock off. Even of those can be made in the system, nobody is gonna release it and if they did nobody would buy it. One last thing...if they are indeed analyzing other peoples works, it could just as easily be to prevent output that actually would be copywrite infringement.
A year ago the best song on Suno, was a complete A.I. joint about a Christmas turkey. It's only when Songwriters/ Arrangers and Producers, (the same talent that creates the goods now) started using the model, the interest, big money investments, lawsuits, and microsoft, rolled in. This thing been out a while. Nobody's interested in that stuff "casuals" put out in any field of endeavor. So all this talk about everybody making hits and blah blah, replacing blah blah, is pointless. Real musicians never made any money on studio work. They'll still make it like they always did--in concerts. Singers --the same thing. The Rihana's are still going to be getting theirs. Studio singers may get even more work because eventually even A, I. users, if successful are going to want a piece of that concert pie and instead of Rihanna, they'll get the studio singer to front their thing. Only thing that is going to change is the people are going to be getting some amazing music again
The irony is that EVERY artist uses this input-output method. We all listen to 100s if not thousands of different artists and genres that inspire us to create our own original musical pieces. This is not that different.
I agree totally. Musicians and artists are always inspired by other artists. 😂 and this is a good thing. You should be proud that someone use your musics to get inspiration.
exactly, all AI does is go many steps ahead and come up with a whole instrumental instead of going step by step as a human would (drums first, melody next, etc. It's simply like having a co/ghost-writter/producer by your side, helping you to develop your idea (I use it exclusively for the upload audio feature), and most if not all of the biggest names have had a whole team of producers and writters or a couple of them at some point.
That's not irony. The human brain learning from creative works, has been part of the social contract ever since the first art and music was made. A computer is not a human brain, and is not subject to the same limitations as the human brain that the original creators expected when first publishing.
Suno & Udio are making it possible for anyone to make music but unless the person is putting in the work needed, the generated songs are going to suck.
Everyone knows Suno & Udio are trained on copyrighted works, but are they simply recreating them or making new songs. The answer is that Suno & Udio are creating new songs based on their knowledge.
Composers fear that AI-created songs will replace them so they want to slow the adoption of AI music to make money before they are replaced.
The record labels want money & power, this hasn't ever changed.
I’ve done some research about trained with copyright works. Finally, even the artists that are inspired by other artists train themselves with copyright works, just hearing it. The fact that they hear a music, they train themselves to learn how to reproduce a certain style. It doesn’t mean that it is a violation of the copyright as long as they don’t make the same exact music.
Yeah but this works like a printer. You could spill out like 200 Beetle sounding songs in minutes compared to one artist coming up with a few songs that have similarities. This crap will be way to powerful in a few years of it doesn't get put in it's place. One this is you can't copywrite this material, however, how do we know if someone is being honest that they didnt record a new song the they made with a prompt (unless all prompt data is public)? @fdupas
@@nwilt7114 You can make songs that sound like any band or artist, the trick is to know how far to go so you don't get sued.
@@nwilt7114 maybe you should try? I bet you won't be able make something similar to beetles in one day, not mention 200 pieces
@@nwilt7114 sorry to break it for you but "honesty" doesn't exist in the music industry, just listen to the most famous singers live, most of them suck and lack either talent or skills, they are fabricated in the studio with melodyne, they have a team of people who are making their songs and many of those people are copying other songs to some extent or using the same sample packs everyone else uses, same for producers like dimitri vegas and those clowns, heck even those you would think were honest like armin van buuren are not.
No one had this tech 10 years ago but you can bet that those fakes from big labels would be using it without telling to anyone if they had the opportunity.
Most of them clips you shown was from Udio... I use SunoAI and write my own lyrics, I love using it as I can hear my songs.. There is no way an Artist or a big record company is going to use my lyrics, I hope Suno wins🥰
Mate we the users are ready to contrubute money and arguments to defend them great
Designers
what is wrong to let AI listen to the radio, it just remembers everything its hearing
I agree, there is nothing wrong. The artists also do that to get inspiration. 😉
It doesn't ingest. I hate the major record labels, they bring nothing but crap to music these days. Hence why I'll use udio or suno to make the songs I used to enjoy people don't make that sort of music anymore and the record industry is only interested in rinse and repeat.
And it isn’t so easy to create a good music even with AI. Finally. What is important isn’t how you create the music but if the result is good or bad. 😊
You make a good point. Popular music today is terrible and these programs give users the ability to make the songs they want to hear.
People can imagine a future where corporations own AI music platforms. They might continue to allow people to create and earn royalties, but through specific terms, they could essentially own all copyright information for the music tracks generated and add them to their own copyright libraries. Over time, people will find that it is impossible to create any musical phrases in AI music that do not infringe on copyrights, as the corporations monopolize every possible combination of notes.
@@holykylin I guess you don’t know that the US copyright office deems any AI output as uncopyrightable and therefore instantly in the “public domain” as soon as it spits out a result. In order to copyright, it’s got to be created by a human. There are case precedents where people have wanted to copyright art made by animals and machines, and they were declared uncopyrightable because a human has to do it. Copyright is for humans only. Educate yourself about this.
@@batautomat Indeed, I am not only aware but also understand that the United States is currently the only country that outright does not recognize personal ownership rights for AI-generated content. Other countries are in a state of ambiguity, or conditionally recognize such rights. However, this does not conflict with what I said. Firstly, the use of generated content isn't just confined to Americans nor is it solely distributed within the U.S. Secondly, while copyright laws related to AI may evolve with the times, private rights are almost impossible to alter. This means if copyright groups legally gain control of AI platforms, it would be nearly impossible to reclaim those rights. Perhaps antitrust laws could intervene? Even so, large corporations often have methods to circumvent these rules.
@@holykylin imagine all you want- it doesn’t mean it will turn out that way. You are using the same method that AI tech bros use, suggesting a future outcome to entice investors but maintaining plausible deniability if they end up being wrong.
I don't think they need to get permissions from copyright holders. Learning is never copyright infringement and transient copies are part of the internet, whenever we listen or watch a piece of media, the browser caches it on our computer, so making a transient copy to read something is generally not copyright infringement.
However if the output is too close to existing copyrighted works, that's obviously a different story.
So is your position: "it's OK as long as they don't get caught outputting something too close to copyrighted work"? (Would you grant similar permission to someone taking things from a store without paying for it, as long as they only used it in certain ways?)
@@daviddas No because if you take something out of a store, it's gone from the store. You're effectively remove something.
But I'd grant a similar permission to everything and everyone to closely examine an object in the store to determine how it was made and then use that knowledge to create a smiliar object in your garage.
@@NorthgateLP fair enough about a physical object removed from a store. But in the digital realm, there's always a transaction when a digital good is consumed. An iTunes music store sale would generate a 99-cent transaction. A Spotify play counts as a micro-payment. A UA-cam view counts as a view (and can be monetized). Because the digital good is under copyright, there's still a transaction that takes place. It's not just a glance at a product on a shelf.
@@daviddas Right, but if the AI companies pay that transaction fee to train their AI then I don't see why it should be prevented from analysing these songs and compose new music based on what it learned.
One important thing that I see a lot of people misunderstand is that when an AI composes something new it has absolutely no access to it's training data and does not contain any copyrighted information in it's model. It creates things solely by what it learned through fancy math.
Now if the AI companies have aquired their training data via illegal means that would be an entirely different story, but then the copyright infringement doesn't really have anything to do with AI.
If someone uses AI to mimic your voice and commit an illegal act, you will deny that it is not your voice, even if it is very similar.
But if the sound is used to make money, You say it's your voice.
And that's exactly what happened.
If AI even further decomposes the copyrighted material in their LLM to even smaller bits like notes would that alter things? You can't copyright notes, just like you can't copyright letters or words. It's a blurry line and I'm pretty sure the big music labels aren't suing because they care so much for the artists...
If they want to ban Suno or Udio, they must also ban all the text to image apps out there, and there are hundreds of them. All of them operate on the same principle, they look at images or songs online and compose new material. I vote for AI to stay, it is the future ♥
You're right, same principle. I don't know anyone who's advocating it to be banned; the only question is compensating the preexisting work that is used to make it what it is.
They should ban themselves to not let the people live their lives how they want. And even the artists are trained with the same technics with active hearing and music analysis, learning harmonies. But at the end, it isn’t sufficient to create a good music, the people need to like it too. If I like a music very much, I will buy it, and if I don’t like, I simply won’t buy it. 😉
if a song has like 5 or 7 consecutive chord progressions that are exactly the same then maybe. But just because a song has parts that sound similar, or overall sound reminds you of anothers does not constitute infringement... if a heavy metal artist does a vocal growl is that infringement?
Sorry, but there within their rights
It is transformative they are not selling The music that they train on if I want to look at a painting that is copyrighted and make my own painting using the copyrighted painting as a reference that is transformative
Sorry, Sony, but you’re wasting your time
I'm concerned that it might still fall under "sampling" in some cases. I hope they will only have to filter out songs which are very similar to existing ones, or pay to copyright holders a little bit if people listen to these songs on their platform.
@@NoidoDev it doesn’t it is completely transformed, if I listen to the radio all day, and I eventually sound like someone singing on the radio that doesn’t make me a thief , and just because technology is used to do this, doesn’t make it any less transformative sorry
The music industry shall fail hard in front of the world by getting told "it is what it is" and "deal with it" and what not.
Cant wait till they figure out how to train these models on less data. Aka only public commons, or self generated and rated dated.
The sooner the ai midels can sidestep the copyright/ip paradigm the better
A problem I see is that it is at the core no different than what people do. People consume music, learn from it taking in elements from different genres and then output music that is often heavily influenced by it. Our brains are updating each time we hear a song. I Is that copying? My guess is that copyright is not going to successfully cover the training because it is technically not copying and we'll need new laws. It will be really interesting to see how this plays out.
Keep us updated sir
You kept saying "Ingested"... I don't think you understand how machine learning works.
Artist artist activists frame these models in a certain way.
Really insightful analysis of this problem, thanks for doing this.
Thank you, David, for this educated video. We all need to be informed and know what is happening in music industry.
AI is a great leveler. Studios, big labels and big artists can't tolerate an ordinary guy with interest in music compete with them. Also every human artist has "ingested" some music which influences what they create. So it's not fair to blame suno or udio. Lets face it. There is no such thing as pirely original music.
To be clear, it's the not the "ordinary guy" they're competing with. It's the AI. An ordinary guy only comes up with a prompt, which is not creating music at all.
This is the process that SUNO and UDIO uses, they train a model, by adjusting a neural net to replicate a close enough version of content based on the prompt given, but the way the neural net stores information, its impossible to ask it to give you precise songs, as it conceptualizes the the style of the music down to the similarities in the musicians voices and composition process. But to actually make the precise songs is like pulling teeth. If you don't believe me, just have Gemini or ChatGPT or read it from Wikipedia how a Neural Net stores information ,its like a fuzzy logic version of nested if-then conditions. Sending input in one end and gaining the results of one complex formula. And there isn't a person on the face of the planet that can identify what any part of it really does, its so complex. And its likely what the Lawyers of Suno and UDIO will provide the music industry lawyers when it asks for an explanation of where the content is stored.
Utilize this technology as a valuable tool to optimize your workflow and improve your output. Also, it is crucial for you to delve deeper into the workings of these LLMs. The level of sophistication behind their operation is truly awe-inspiring once you grasp the intricacies of these systems. They possess a remarkable ability to emulate the human brain, enabling them to learn from information and generate novel concepts. These companies that are filing lawsuits simply want control, whether due to a lack of comprehension regarding how these systems operate or a desire to profit from them, or perhaps both.
I look forward to how this plays out. I am under the impression that the way the companies trained their data is fine. I have a band, but I also write a lot of music. I think the ones who are worried about it are those who are composers, record labels, fill in artists ect. I am a singer who has a bad voice, and in my band I use effects and things to make it sound okay, I have used auto tune for years, and also beat randomisers, and other tools to get ideas. My fav tool is the band in a box pedal, but of course I don't get the beats from the pedal, once I have a functioning idea, I record it, then do the drums, bass, vocals around the guitar. With ai music generation, I have been able to get the syllables of my written lyrics, and prompt an ai gen with what I want. Then I work on that. I was also in a band with a lead singer who sounded like Shirley Manson, we didn't do garbage covers, but some people thought we did.
Well the reason why the lawsuit shouldn't really frighten anyone is the same reason why other platforms who do similar things shouldn't be worried either
(or even be fearful of a similar lawsuit).
For Example; Twitch implemented "Famous" voices as TTS voices for donations to streamers, trained on song artists voices, and mostly all (if not all) are trained on song data - The Weeknds voice is completely trained on one song; if you wrote down the lyrics for the song in your TTS, the song would essentially be played.
Although the data is trained on these copywritten voices, anyone can get paid with the addition to these voices, adding formative content to their streams.
Where is the line in this regarding copyright , and how does it relate to infringement as a whole for AI training data? well that's the whole point of the example, these things still need to be defined, and lines need to be drawn before estimations of crossing those lines can even be made.
In the same way these companies are mad that these AI generations are creating things using previously existing art/ "their art" - the logic should fall under all matters of this magnitude, including but obviously not limited to that example.
The problem with that, is that the very foundation of the lawsuit
(This is data trained using our licensed music)
Needs to first have a category to even fall under the guidelines of what is considered "wrong" of them to do, *first*, before they are even considered for doing "said wrong thing" within the confines of those guidelines.
The courts need to define what they consider within the fair use-age regarding training data for AI as a whole.
And THAT is a process that I don't think these companies realize just how long it is.
Even if the AI companies are in the wrong, they first need to be defined within the confines of what type of copyright infringement the Music industries are claiming first. They might not even get that far.
And this will take at the very least a decade to even draw the line of distinction.
All in all, I believe this won't get very far. By the time they draw the lines on AI training data to even start this fully, the lines will be even further blurred.
It's tempting to view these lawsuits as a David and Goliath battle but when an AI model creates songs that don't just sound like the Beatles but sound like Paul McCartney singing on them then I can see why the major record labels may have a point.
I think what labels really want out of this is to protect their sampling businesses. If you allow the theory put forth by suno and udio to proceed without challenge, then transformative use, not just ai, but all of it, needs to be reevaluated, and they lose massive amounts of royalties they're making today. This might seem like it's about ai, but it's really about understanding the rules of the road, going forward, at least for the next 30 years or so. I'm glad they've opted to sue the corporations and not individuals though. Given the RIAA's history with this kind of thing, I'm very happy that they're going to a place where this can actually be resolved. No matter what happens, or who wins, ultimately, a decision here, any decision, would benefit everybody
A lot of music done with Udio and Suno can be awful too if you dont’t create good prompts
Yes...but left to its own devices, it will get better in a hurry. Imagine where it'll be 6-12 months from now.
@@daviddas there are also very good musics done with Udio. I'm also using it and I love it. I don't understand why they always complain about the fact that everything is trained with something. Even people are trained with something when we study, so what's the problem to use AI ? For me, that can be very helpful in everything.
If you can hear the song in your head and recite/play it, your brain made a copy of it. Treating AI differently when it comes to the process of learning through experience is (at the core) a human bias that creates the precedent that learning and performance is illegal.
Great video, as usual! Thanks for laying it all out, David!
Great Video David! We've created our youtube music channel few months back thanks to Suno and Udio. These AI platforms gave us a chance to share our creativity, even on Spotify. Something we couldn't possibly afford doing before.
I usually am not a fan of Gen AI, HOWEVER, the music industry is a plaque anyway. Michael J. and Prince already mentioned that. In this case, I actually would band with gen AI platforms, just to annoy those big wigs.
artist before AI were already making music that had hints of other artists. based on older music etc, soundsbits.
Exactly 💯
Atleast you music guys have these corporations fighting on your side, in the visual art space shit is ruined beyond repair. Mid journey has completely disrupted the entire industry.
Thanks for this video, David. I appreciate your dedication to exposing the nuance in a situation that from the beginning was intensely polarising, and remains so today. Please do continue talking about AI and its relationship to music so that we can get your take on any further developments in this space.
May be a hot-take but I think ai models should be allowed to use any material they find on the internet as long as that material isn't directly included in the product result. This is what we humans do, we take inspiration from any and everything we can. I'm allowed to look up a picture of the mona lisa, study it, and lit that knowledge influence my artwork. Hell, I'm even allowed to paint as near of a replica as I can manage. AI is designed to learn in a very similar way to how we are. It will outperform us in every way eventually, slowing it down doesn't really change anything.
A bunch of rich people are looking for ways to cash out on some abstract controversies, nothing new.
If this lawsuit is successful, it means all ai tools will have to be pretty much rebooted as they all borrow and learn from all available material. Suno has already hit back at this pretty hard.
I'm also wondering what ai tools Sony, Universal and Warner are using. If they use any at all, it would be good to point that out.
If "kinda sounds like and reminds me of" becomes a copyright violation.... kiss the act of making music outside of the permission of the big record labels goodbye.
This is a well-trodden path and not particularly clear in law or the courts (see Blurred Lines).
God forbid something that a major company doesn’t own exists. The lawsuit won’t help any musicians. This will only put more money in the pockets of the greedy people that ruined music decades ago. I’m mean seriously when is the last time you turned on a radio.
Because of A.I music, music is amazing again, they just want to profit off of AI Music smh. Fuck the music industry and the labels
Great discussion
I want to thank you.
All music created is based on previous works essentially. As a musician I grew up studying and learning music and those influences absolutely are present in the music I create. The only issue here is Industry Rule #4080, Record Companies are shady. Embrace the tech. The best music will have human input. Prompts are no different than a producer telling session musicians what to play.
If you say that "users of Suno and Udio aren't making musing, they're entering prompts", then you could also argue that "the person who plays the piano isn't making music, he's pressing keys". I understand that AI is heavily disruptive to the industry of creators and distributors, and that society might want to think of a way to protect those who could be harmed severely by that. (I would support that idea a lot.) But the current arguments to sue software companies for their technological innovation seem quite ridiculous. You can easily copy a painting by taking a picture of it, but until now, it was always the responsibility of the photographer... and when trying to sell that picture, he should be held accountable if that would be a copyright infringement - nobody would think of suing Canon or Nikon or another camera manufacturer. Why is it so hard to treat the current audio creations in the same way? When a song created by one of these tools is considered a copyright infringement by a court, then the person who tries to make money from it should be the one to blame.
See that?...what?...there's a tsunami of creativity headed our way!
They ban others from using it, but they will use it. They got big appetite to feed themselves. We all know that. I’m using Suno and Udio, they are good. Traditional music producers need to grow and move along. This is the next phase in music creation. Hate it all you want but this is the next direction, or be left out. It is developing and learning, and it is growing and getting better by the day. It’s a hit and miss to get what you really want but you do get nice ones. It can be used to find ideas and re-record nice parts, and different parts together to get a full new nice song.
I hear your arguments as a composer and I respect them. As someone who found Sumo this week, I;''ll add this. I'm a wannabe fiction writer so I respect the position of 'Creatives. I've also fancied writing lyrics but I have no composing ability... at all.. I'm not just using prompts in suno, I'm writing my own lyrics, or changing the ones it creates so its 75-100% my words... Then generating the tune and performance. If i take my work anywhere outside the 'Suno community' and my friends, I'd like to think I could find a composer who could write new music or arrange what I've created... But these are early days of my relationship with Suno, so who know?
Well-known songs and artists are imitated by stock music composers/producers who tag their tracks with genre, mood, style, BPM, and sound-alike keywords/names. If a generative model was trained on that data, it could output music that sounds very similar to the original well-known work without ever having directly ingested it.
Thank you David...your video was AWESOME and you explained it very well. I teach news related English in Korea, and this was a story we did in class yesterday. I tried my best to explain it but I shared your video with my students as your explanations are perfect and clear with some great examples!
Fantastic video!
Cheers! 👍👍👍
If we look into history we find that many of the real masters behind the real artistic abilities were many times caught into contracts and were never really paid what they were worth.. In addition some of songs we hear and know today were stolen and made popular. So my view is it was never about protecting the artists, it was always about protecting investment, and margin and generating turnover. Many of the artists have aired those facts over the years and I can give specifics... Look at the classic composers of generations past, be it music or art.... A lot of that music and art was never paid and the artists died poor or from sickness but their works sell for millions today. In addition any type of music has rules built into it and every dictionary is limited . Music is not unlimited as many may think because the life experiences within a lifetime are recursive in each new generation....These are the subject matter for any good music... So the record labels needed to see this change. What they have done to other artists are now happening to them in exactly the same type of odd way. They are getting cut out of the picture and I find it amusing that they are complaining about that.....The solution however is to find a better way for artists to share their music without the bigshot labels.Take away that waterballoon and probably the music will be affordable and the artists would earn what they deserve, so that the revenue goes to the artists and not to the labels..... We use AI today to find answers we dont have... as a helpful tool... Did we not think that one day someone would come along with ideas to tip the balance back into the favor of common folk ? Supporting and enhancing the ideas of artists that would have remained unknown to the big labels. I also however draw the line that the real artists should get a piece of the AI pie if it learns / feeds on a library of already made music--- As for the labels you can make anyone into a star just by having the right tools, however taking music that we all love, live and work by, it being more a heritage, and then holding that ransome by moving it into a money making business, also means that those who cannot pay dont get to hear it..... Or artists that dont pay the big labels will never make stardom... I´ll leave with another thought, "How long will it take the AI companies before they decide that their tool makes them part owners of the art / music in the future.... Maybe we should find the person that made the paintbrushes that M.Angelo used to paint with and pay him billions for part of the sucess story.... Maybe someone can create a song about reoccuring life experiences and put a patient on that...... as if no one else would ever have that experience or come to any of the same conclusions..... ;)
I think it’s funny lol. A.I don’t have any idea of what is actually is excepted. It follows commands and what ever you feed data wise into it. Udio can be changed and made where the voices don’t have to imitate a existing artist. Maybe parameters would be useful
Incredible that this techies even ADMIT that they completely ignore EVERY SINGLE LAW that (at least marginally) protects working artists like myself. And what is most depressing? I have plenty of friends/colleagues/etc who use AI art and music frequently, for work, for pleasure, f.e. kids birthday and so on = it is a self-destructive cycle!
Record companies are finally reaping what they have sown. For years, they have exploited artists, profiting immensely from their talent while giving little in return. However, the advent of new technology heralds a transformative shift that could drive these companies into bankruptcy. Their arguments against this innovation are utterly unfounded. Artists have always learned from one another, drawing inspiration from chord progressions, rhythms, and harmonies. Now, AI enhances this creative process by providing us with prompts that lead to original work. There is no fundamental difference between the ubiquitous use of the same four chord progressions in countless songs and the use of AI in music creation. I find it amusing that the record industry, after years of exploitation, is finally facing the consequences of its actions.
How is this any different than image ai systems trained on copyright images? Or text systems trained off scraping the web and other copyright material???
It's exactly the same, and the other gAI models are facing the same questions.
You're right on the money. There is a mistake in saying that these aplications "listen" to music or they are "trained" by music. So later on they imitate those recordings parameters. Becase they're not living consciousness. They don't listen to music, they incorporate those recordings into their databases and use them in order to operate. That is commercial use of a copyrighted work without a license. Copyright infringement 101. We don't need more laws, we need to enforce the copyright law that is already In place and ban these applications for good. All those companies would be out of business if they are obligated to pay just 1% of the penalties of the violations they've already done. And there is no safe way to let them be because the ripped off has already been done on a massive scale. And I agree with you that the major labels don't want to eliminate those so creators can make a decent living on our creations. Otherwise they would have never give away their masters to the streaming platforms. They were the ones sponsoring the streaming monster while telling us that Spotify is the bad guy. (I'm a songwriter)
That's not how they work. Effectively they listen to the songs. The songs are input into the machines and the machines update their parameters to predict what's going to come next in the song. And they are trained on a large volume of songs and end up learning parameters that can generate a large variety of music. It is "training" and not really so different from what people do. Of course though like with expert musicians, if you tell it to reproduce a certain style it will be able to. That is the violation of copyright and would be a violation with people as well. Considering the training of the parameters a violation would have numerous potential consequences that are not clear. Don't get me wrong though. I am not arguing for it or against it.
They don't need to be conscientious. They don't have the exact music in their model, it's only a approximation.
@@ssssssstssssssss They don't have ears, so no, they don't listen. They record, copy or scan audio files and incorporate all that into their database. the softwares are "Using" copyright material to operate without a license. that is copyright infringement. The trick is to fool the people making them believe they are living entities so they do the same as we humans do. No matter what wording you use the outcome is the same. Feeding their databases with all of our works in order to operate. Using copyrighted works commercially without a license. And all of this is agravated because it has been designed to put the creators out of business to push political agendas. #1-The Digital ID, #2The Universal Basic Income. This is criminal against humanity. Besides if you want to be more credible change your profile, make it look a bit like there is someone behind not a bot or a sponsored troll.
@@NoidoDev They operate using those copyrighted works so they can spit those approximations you mentioned. No matter how they use those copyrighted works, they're using them without a license. Copyright infringement 101. But even those approximations you mention are so poorly done that you can easy recognize the songs they're copying and in some cases they even copy exactly the same lyrics of famous songs exposing themselves as plagiarism. If the law is enforced they will be out of business inmediatly. But there are political agendas behind these Ai apps.
@@DanyTomaslifehave you ever used any of these ai programs? I cant tell anything about udio but i have experience with suno. I spend hours just to make a song exactly how i want to sound like, and no, it never sounds like any music i listened to in the past from real humans. Congrats for taking away the chance from poor people to have some fun and create their own music for themselves. I know it has to be regulated somehow, but the sad thing is that the big music companies want their profit and position, or if they really cared about the artists we wouldn't have such a horrible music industry. That's why I started to make my own music for myself. I am not going to support something I am not satisfied with, aka the shit I can hear on the radio again and again. I am gonna support what I like, and I am going to listen to the music I like. Suno made it possible to me, period.
7:48 What about samples???
Read the terms and conditions of Spotify or Apple Music, or the back of any CD and it says "Non-profit use only." Training AI on copyrighted music in order to sell the music it creates is for-profit. Ergo, all of Udio's output is in breach of copyright. They need to be sued out of existence!
Isn't this literally how Spotify was created? The creators needed music so they took it without permission and got forgiveness AFTER the fact, and now the major music labels are part owners of the platform. So don't be surprised when these labels end up partial owners of these AI platforms. Suddenly they have their "cut" and aren't mad any more. Meanwhile artists will get next to nothing.
When these companies get a piece of the action this will be solved
The fact that I can show you or anyone an entirely AI generated song without them having any idea it was not human made disproves anything you can say about the difference between AI and humans. We are literally there. AI can be entirely convincing as an actual person.
When machine can mass produce output, commercial value of the human talent decreases. That's the way it is. This is the new reality for musicians.
See they cant sue because they arent profiting.
I'm with you 100%. 'Fair use" NEVER meant 'for fair profit'. Educational and news commentary, and only in segmented clips is what Fair use is meant for. I hope these guys are sued to oblivion. If anyone can push a button and get a competent song, then music becomes worthless. AI is strictly for untalented people. It's bad enough with spotify giving anyone access to all the music in the world for eleven bucks a month and now we artists have to deal with this?
I got a few minutes in to this and up pops an advert for AI generated basslines! It asserted that back in the day when people started using laptops and software people using big consoles and tapes said that was cheating. I don't remember that. Computer stuff was actually alot more expensive and the evolution was analogue, then these huge digital systems for hard disk recording that only major artists could afford. Everyone else still used consoles and tape until the costs of digital came down. No one said it was cheating.
Interesting. It wouldn't surprise me that advertising would target AI solutions for musicians.
The thing is any human can copy a song and it is humans that are creating these similar songs.. The law should not be based on how the AI learns it should be based on how the end user uses it. There needs to be more safeguards to prevent direct replication of songs whether it be instrumentation or vocal melody or likeness of the singers voice while still retaining the knowledge of previous work to inspire future music. Exactly as humans have done for centuries.
I have plugins in my DAW which handle vocals I’ve written and another ones to suggest bass and drum lines after analyzing singer’s voice. After that I can sidechain other plugins to generate guitars and strings, etc… Yet this is considered to be my own composition. Having tried Suno, it prodeses soulless radio formats without real knowledge of song’s meaning. Sort of nice to try if the genre you thought of works, but it’s still less stressful to have real musicians singing and playing them as they are intended.
As a musician, I spent my life never listening to any music - only writing and performing my own. Otherwise I might have been influenced by that music and broken copyrights. Maybe that contributed to my complete lack of success as a musician.
Spot on. It's just a fancy 2024 way to rip off artists. "instead of entertaining a good faith discussion, they've reverted to their old lawyer-led playbook." Yeah, well Bubba, that's what labels do, they protect their intellectual property.
The funny part is Udio and Suno said they "care about the artists, that's why we ban artist names in our prompts", when they are doing it to hide the fact they've trained it on copyrighted material. Practically every song on Spotify is copyrighted.
Regarding your final comment on everyone getting a share of the AI music, that only applies to artists on those labels I'm sure. The only good outcome for indies is these companies will be stopped. Anything else will still mean a cheapening of artists' work, ie the prosecution's argument will most likely still apply to any artist not already on a label.
Great video, as a music and video creator I was wondering when, and hoping for, the day these AI companies get taken to task for nothing less than outright theft. As you say, they are not making 'musicians' either, it's just a load of talentless people typing prompts - it's only going to homogenise music not be creative. I wonder when the big image companies (say Getty) will also take action. What should happen imho is the AI companies should only work from a licenced pool of music/images - maybe artists can freely upload into those pools. When ever any of the music or images in that licenced pool is used (and they can easily track that) for any AI then the copyright owners/performers etc get credit and a financial proportion of the AI company profit. This will mean AI won't be free, but it shouldn't be if it's just is parasite feeding off others people's creative endeavours. We had the same sort of thing when mp3s were being generated and shared for free, it seemed impossible to control at first but once the law cases kicked it it got changed.
You're right. There is no problem with this technology if the data foundations are legitimately created or licensed.
Udio isn’t going to take your composing jobs, Sony, UMG, and the rest will. Udio users won’t replace anyone lol. They busy making music about farts or cats.
It's only a way for the majors to get a peak at suno and udio's code... to be able to do this by themselves!
Interesting conspiracy theory. :) And there may be a bit of truth in that. Eventually these algorithms will become public, copied, and commonplace.
The second artists start licensing their style and works to be used, every song will be just another copy. Everyone will be making their favorite artist's work. Originality... gone. The protection so to speak that Suno and Udio have in place that blocks the people who, as you put it, prompt, is there to keep people from doing that and it keeps the work original. When some label leases out the rights to like a Billie Eilish song, every song is You Should See Me In A Crown.
Oh.... Music is music regardless. Music can't mimic music because it is inherently that.
What about the famous mainstream music artists of the past who stole from Black music artists?
I think the vast majority of these would be classified as inspiration. If you have specific examples of "stole"n music, i.e. theft, you should post specifics. Every musician who has ever lived has been inspired by artists who came before them, whether they be from their own tradition or another.
What is the difference between an AI model ingesting any recorded music, data on music theory and an artist spending days, weeks or months listening to the same music and reading endless articles, books etc on music theory and sheet music a being influenced by it to write, record and perform something similar?
My view why these record labels are only going after Udio and Suno rather than other players in the AI music generation market (even so called ethical players) is because there is a difference between them. The likes of Soundraw, Soundful and others only produce audio output and backing tracks at best because you don’t have any real control over the output, but Udio and Suno standout because they can produce the audio in a way that is more creative and a kin to actual songwriting by using meta tags for thing like intros, instrument solos, voicing styles. Add to that the addition of vocal output then it is a danger to their business, some would say to artists as well. But to me an amature creative with some musical skill and ability to write lyrics they provide a tool to produce something that not so long ago I could only hope to achieve.
I also find it strange they are not suing the AI voice changers in the same way, as these provide an equal threat to them and their artists because they allow recorded vocals to be swapped out for somebody else who didn’t actually record the vocals, but maybe the reason for this is that they are actively investing in these voice changing AI models so that they can exploit them and their artists by being able to create new Beatles songs or have Billie Eilish cover a song she doesn’t want too.
I think it's great that AI creates music that "sounds a bit like someone else's music", but how do humans use this song? The moral standards are still firmly in the hands of humans. If the similarity reaches a certain standard, it will be labeled, just like the rating of movies.
I understand why song writers and musicians are bothered by this but I don't think there is much that can be done about it. Automation is going to continue to replace the need for humans to do certain jobs. I do not think A.I. generated music will ever fully replace human creators because we all like genuine talent but it will compete.
@ 4:48 Mikey Shulman saying "Suno's mission is to make it possible for everyone to make music.
It’s like he saying “ My Wife is for everyone anyone can butt f**k her.” It’s just like that 😂
see my problem is that i have specific type of music that i like and its basically impossible to find good artists making those so im going to still make those myself and share them
This lawsuit looks like to me a way to shutdown the competition that the big record labels can’t no longer compete.
I completely agree if the content is copied, the creator has to be cited, compensated, etc.
I use Suno, but I write my own lyrics. The songs crated by Suno with my own lyrics it was never copyrighted.
There we go! You mentioned about AI replacing composers like you, maybe you should try to incorporate AI to your advantage and keep up with AI. AI is here to stay.