Neural networks violate the rights of musicians
[ad_1]
Musicians united against the abuse of neural networks. The American Alliance, which fights for artists’ rights, has published an open letter calling for a ban on the use of artificial intelligence in the industry. The main complaint is that such technologies infringe on and devalue the work of authors. According to the musicians, the developers illegally train neural networks on songs and do not pay royalties. The topic will be continued by Grigory Kolganov.
Bon Jovi, Stevie Wonder, Katy Perry, Billie Eilish, Imagine Dragons and over 200 other artists spoke out against the illegal use of neural networks. Musicians are asking companies to stop training artificial intelligence on their voices and works. This is primarily a matter of money, developers do not buy the rights to songs, and artists lose a significant part of the royalties. Videos with digital copies of the voices of real musicians are multiplying online, a phenomenon known as AI Cover. It is not easy for an untrained listener to distinguish generated vocals.
In an open letter, show business stars indicate that songs generated by neural networks are sabotage and undermining the reputation of real artists. At the same time, performers acknowledge that the technology has “huge potential” and can inspire something new. So, last year, the neural network completed the last song of The Beatles, called Now and Then, in which the real voice of John Lennon sounds. A similar practice exists in Russia. For example, the group “Technology” rethought the composition “Strange Dances” from 1990. There are no live vocals in the new version.
The new version was generated in just a few days. Producer and member of the Tekhnologiya group Leonid Velichkovsky told Kommersant FM about the procedure: “The specific task was to recreate the voice of Roman Ryabtsev of the early 90s, those emotions, that timbre. We are all not getting younger, but getting older, and our voices are changing. The creation took several days, that is, Roma’s old voice was isolated, the neural network listened to it and then sang in the same voice, 90% of whom everyone really likes.”
As for the legal and economic aspects, as experts interviewed by Kommersant FM note, proving the illegality of using the compositions of a living artist is not an easy matter. The voice is not protected by copyright, which means legal proceedings may drag on. Such impunity gives rise to many abuses, the listener should exercise maximum vigilance, summarizes the head of the information security department of SearchInform Alexey Drozd:
“In 2023, scammers began making money from deepfake music. They took the rapper’s tracks, faked his voice, forced the neural network to voice the lyrics, and generated similar music. They put demo recordings of new songs stolen by a hacker on the platform as an unreleased album, that is, they created a veneer of exclusivity. And it worked, I think they were able to earn $80 thousand.”
Streaming services have been releasing stylized tracks for years. For example, Yandex programmers created the albums Model and Neural Defense. The lyrics were written by a neural network in the style of Kurt Cobain and Yegor Letov.
Everything is clear with us – Telegram channel “Kommersant FM”.
[ad_2]
Source link
