How AI will change the sound industry

Why the ethical use of neural networks to create music is almost impossible. The journalist and musician Anton Odino are sorting out

Recently, more than a thousand musicians, including Hans Zimmer, Kate Bush and the Jamiroquai group, released the album "Is this What We Want". It included 12 records of silence in studios and concert halls. The names of the tracks are formed in the sentence “The British Government Must Not Legalize Music Theft to Benefit Ai Companies” (“The British Government should not legalize the theft of music for the sake of enriching and companies”). The album is a protest against the initiative of the UK government to allow the AI ​​developers to freely use copyright, unless the copyright holder prohibits it himself. Suddenly, the British authorities went to meet the artists and agreed to revise the bill.

Journalist and musician Anton Odino tells how neural networks change the musical industry and why they can become a big problem.

Why and companies works of art?

In itself, artificial intelligence does not know how to generate text, nor images, nor music. In fact, all that he is doing is mathematical calculations, and the task of its developers is to achieve the necessary results from AI. For this, the neural networks give a set of tasks and ready -made solutions, and it should develop algorithms itself, which allow you to come to the right answer. The more attempts, the more accurately the result, and for more attempts you need more data. And here, the developers of AI have a need to use someone else's content.

To teach ChatGPT and its analogues to generate a coherent and meaningful text, they need to “feed” the terabyites of examples. The same with images, music and any other content. If you put it very simple, the neural network cannot create something that she was not shown in advance . The developers of generative neural networks, as a rule, do not disclose what data they teach their models, however, projects such as the recently presented Openai picture generator in the style of the Japanese director Anime Hayao Miyazaki indicate the use of works protected by copyright.

In itself, AI as a technology is neutral and does not violate rights. But companies working with this technology, and especially the creators of generative projects (Openai, Suno, Udio or Midjourney) are increasingly ignoring the concept of intellectual property. And most likely, without this, their services could not develop so rapidly.

This is a completely new legal conflict: several claims are considered in the courts against companies developing AI, but there have not yet been made decisions on them. That is, there is no legal precedent that would determine the rules for licensing content for teaching neural networks. AI-companies remain in the "gray zone."

The richest child in the world

The problematicity of generative and startups can be considered on the example of the Suno music generation service-both from the legal and philosophical point of view.

In the summer of 2024, the American Association of Soundwriting Companies (RIAA) filed a lawsuit against the company, accusing the copyright violation project. The startup used other people's songs to teach his AI models. Suno in response was recognized that, indeed, they trained a neural network on tracks protected by a copy, but this was a “good use” . So in the USA it is called a legal doctrine that allows the use of works without the permission of the author, if this contributes to the "progress of science and useful arts."

Responding to RIAA's claims, the founder and general director of Suno Mikey Schulman compared artificial intelligence, who is trained on music available on an open Internet, with a child who tries to write his own songs, having heard a rock. Schulman agreed that most of the material was protected by copyright, but emphasized that he did not consider training a violation.

There is one significant difference between the child from the example of Schulman and the Suno artificial intelligence model: the service subscription costs $ 10 a month. In an interview with Rolling Stone, the founder of Suno voiced a plan to attract a billion subscribers, that is, to reach an income of $ 120 billion annually. The richest musician according to Forbes in 2024 was Jay-Ji, his fortune was estimated at $ 2.5 billion, in second place Taylor Swift with $ 1.6 billion. That is, if everything goes according to plan, the “child” of Shulman will earn ten times more than the two most successful artists in the world combined.

In the podcast, Rolling Stone Music Now Schulman carefully leaves answers to the questions of the presenter about the ethics of his company. How does Shulman relate to the fact that business prefers to use the generative tracks of Suno, instead of contacting living musicians? The startup says that he dreamed “so that any gastronoma had its own jingle,” and now thanks to Suno they have such an opportunity.

The problem is that small business could afford their own music before, notes the composer and researcher of AI Ravi Krishnasvami. A jingla recording of a living composer costs from $ 10, if you order it through the Fiverr freelance tour. Perhaps the lack of gastronomas of their own musical topics is not connected with the fact that before Suno they could not afford them; Most likely, they simply do not need jingles.

But Suno can really replace the work of part of the musicians, especially those who earn the creation of functional tracks for audio -bit, podcasts or advertising. The same applies to representatives of other creative professions (illustrators, artists, designers), whose work is not a very demanding customer to prefer and generation. Studies on how neural networks overlap the labor market are published regularly. So the meme that designers and creators of the jingles will be replaced by AI is no longer a joke.

Colonization of knowledge

Shulman rests on the availability of music with Suno. “Technologies make the future possible in which anyone can explore, create and actively participate in the area of ​​art, where the majority used to be consumers,” he writes in the same post with a response to the claim of RIAA. At the end of 2024, his quote that "most people can not be pleasure, the process of creating music, was clarified (mainly by the forces of indignant musicians).

The founder of Suno also insists that in creating art, taste is more important than skill. However, in the process of teaching neural networks, the taste of developers who select data for AI plays a decisive role and evaluate the quality of their answers. But this contradiction is the smaller of the problems with generative AI. It is interesting to look at this technology in the context of modern neoliberal capitalism.

The Canadian researcher Nick Snrnichka calls the economic model that has developed after the financial crisis of 2008. By platforms, it implies digital services that act as intermediaries in the interaction between different groups of people: sellers and buyers of services and goods, advertisers, employees and employers.

We can talk about almost any famous online services, whether it be Facebook, Amazon, Uber, Spotify or the mentioned Fiverr. Indeed, for the most part, they are convenient for the final user. But the reverse side of this convenience is the absorption of entire markets with such corporations, bypassing laws and state lobbyism, employees' insecurity, opaque collection of user data and their trade (they can be used in machine learning to automate work and saving people).

We recommend that the proletariat 3.0 be bookmarked : why we work, but we do not earn anything inequality capitalism date of publication 03.28.03

The mediation of the corporation is turned into assets of the phenomena that previously belonged to society, whether collective knowledge, communication channels, infrastructure or data, the researcher of the Canadian York University Kein Burch draws attention. An example can be streaming services, such as Spotify or Yandex music, which provide access to protected copyright music, earning on a subscription or at the expense of advertising (and paying the musicians for the most part miserable royalties). But this concept is especially accurately described by AI projects.

Everyone who has access to the Internet can now write songs, poems or paintings - it is enough to pay for the subscription to the corresponding service. The most tiny business can afford its own music, visual style, text content and even automated communication with clients. Knowledge, which was concentrated in the hands of a large, but limited circle of specialists, is now available to everyone. This could be called democratization, if you take out how this accessibility is ensured by the brackets.

In February 2025, the court found that META used more than 80 terabytes from the Libgen pirate library to train their language models Llama. In response, the company said that it was not shared by downloaded files through the Bittorrent protocol, and therefore did not formally violate the law. Openai appears to the same arguments about conscientious use as SUNO, while lobbying for permission to use the materials protected by the copy in order to preserve the “technological leadership of America”. The court decisions on numerous claims on the violation of copyright against the AI ​​have have not yet been issued, but with the current introductory we can already talk not about democratization, but about expropriation or, if you like, colonization of knowledge.

Against the background of these proceedings, the story of Aaron Schwartz, an American Internet activist, a fighter for freedom of information, is indicative. At the end of 2010, Schwartz downloaded 70 gigabytes of scientific articles from the base of the Massachusetts Technological Institute JSTOR on his laptop and wanted to lay them out in open access in protest against inequality in access to academic publications. He did not manage to distribute the materials - on January 6, 2011 he was detained by the police.

Although the institute did not file a lawsuit against Schwartz, federal prosecutors seriously took up the activist. In July 2011, he was charged with several crimes, including computer fraud and damage to property, which he was threatened with up to 35 years of imprisonment and one million dollars of a fine. The process dragged on for more than two years, and in January 2013 the activist committed suicide.

Technological (anti) utopia

Platform capitalism really simplifies life. For almost any task, there is a service that will help to cope with them quickly and at an affordable price. People who directly solve these problems are either invisible and functional or formally equal to the client.

AI-revolution goes even further, completely freeing a person from work, both routine and creative. Many needs can already be closed either by pressing one button, or using the correct industrial plate . The rhetoric of and visioners captivates about endless new possibilities-in the end, I just want to dream of such a future.

In a neoliberal paradigm, when everyone is an entrepreneur and each for each competitor, artificial intelligence, if it can free someone, then only a business from unnecessary costs. However, the technology already exists, and it is unlikely that the development of new models of AI and the destruction of the data centers, as the famous technopessimist Elizer Yudkovsky proposed , will help to rectify the situation.

Nick Snrnichka, for example, urged to create his own utopia, an alternative neoliberal model of social relations, which at the same time would welcome technological progress. This idea, although quite abstract, is much more interesting than the last proposal of Openai general director Sam Altman - to purchase more graphic processors so that Chatgpt generates the pictures even faster.