BPI Calls for AI Labels on Spotify and Other DSPs

AI

Digital Music News:

“That’s why we’re calling on the UK government to protect copyright and introduce new transparency obligations for AI companies so that music rights can be licensed and enforced, as well as calling for the clear labelling of content solely generated by AI,” Jones indicated.

The latter adverb raises interesting questions about what an across-the-board labeling system would look like in practice. For obvious reasons, if they do incorporate AI, established artists probably won’t want their music labeled as such – hence the “solely” clarifier.

This seems like a no-brainer to me.

Probable ‘AI’ Artist Gaining Spotify Traction

Digital Music News:

Has an AI “artist” topped 474,000 Spotify monthly listeners (and counting) in just weeks? It sure seems that way, and the development is raising new questions about machine-generated tracks’ streaming prevalence.

Word of that presumably AI-powered “band,” The Velvet Sundown, appears to have started circulating earlier this week on Reddit. There, multiple users said some of the relevant tracks had arrived in their Discover Weekly playlists.

🤮

Record Labels in Talks to License Music to AI Firms

AI

Lucas Shaw, writing for Bloomberg:

Universal Music Group, Warner Music Group and Sony Music Entertainment are pushing to collect license fees for their work and also receive a small amount of equity in Suno and Udio, two leaders among a crop of companies that use generative AI to help make music. Any deal would help settle lawsuits between the two sides, said the people, who declined to be identified because the talks could fall apart.

AI Generated Music Is Hard for Humans to Detect

AI

Deezer reports that around 18% of new tracks uploaded to their platform are AI generated:

“AI generated content continues to flood streaming platforms like Deezer, and we see no sign of it slowing down,” said Aurelien Herault, Chief Innovation Officer, Deezer. “Generative AI has the potential to positively impact music creation and consumption, but we need to approach the development with responsibility and care in order to safeguard the rights and revenues of artists and songwriters, while maintaining transparency for the fans. Thanks to our cutting-edge tool we are already removing fully AI generated content from the algorithmic recommendations.”

Meanwhile, a new study says humans aren’t great at detecting AI generated music:

“The average score was 46%,” O’Donnell reveals. “And for a few genres, especially instrumental ones, listeners were wrong more often than not.” O’Donnell says when he watched people take the test, he noticed that qualities they confidently flagged as AI compositions—fake-sounding instruments, weird lyrics—were not always right.

“Predictably, people did worse in genres they were less familiar with; some did okay on country or soul, but many stood no chance against jazz, classical piano, or pop. Beaty, the creativity researcher scored 66%, while Brandt, the composer, scored 50%.” With just a few text prompts, O’Donnell created music that humans couldn’t pick out of a line-up as AI generated. “A few could have been easily played at a party without raising objections, and I found two I genuinely loved, even as a lifelong musicians and generally picky music person,” he shares.

YouTube in Talks With Record Labels Over AI Music Deal

YouTube

Financial Times:

YouTube is in talks with record labels to license their songs for artificial intelligence tools that clone popular artists’ music, hoping to win over a sceptical industry with upfront payments. The Google-owned video site needs labels’ content to legally train AI song generators, as it prepares to launch new tools this year, according to three people familiar with the matter.  The company has recently offered lump sums of cash to the major labels — Sony, Warner and Universal — to try to convince more artists to allow their music to be used in training AI software, according to several people briefed on the talks. 

Explaining the RIAA’s Lawsuit Against AI Music Startups

Legal

Devin Coldewey, writing for TechCrunch:

Like many AI companies, music generation startups Udio and Suno appear to have relied on unauthorized scrapes of copyrighted works in order to train their models. This is by their own and investors’ admission, as well as according to new lawsuits filed against them by music companies. If these suits go before a jury, the trial could be both a damaging exposé and a highly useful precedent for similarly sticky-fingered AI companies facing certain legal peril.

The lawsuits, filed by the Recording Industry Association of America (RIAA), put us all in the uncomfortable position of rooting for the RIAA, which for decades has been the bogeyman of digital media. I myself have received nastygrams from them! The case is simply that clear.

Scraping the Web Now, Asking for Permission Later

Apple

Federico Viticci, writing at MacStories about Apple’s details on their AI model being trained on web content:

As a creator and website owner, I guess that these things will never sit right with me. Why should we accept that certain data sets require a licensing fee but anything that is found “on the open web” can be mindlessly scraped, parsed, and regurgitated by an AI? Web publishers (and especially indie web publishers these days, who cannot afford lawsuits or hiring law firms to strike expensive deals) deserve better.

I agree wholeheartedly. I felt similarly when I looked at the data that trained Google’s AI. I see Chorus and our forum very clearly in their training data. We didn’t agree to that. Our community never agreed to that. Google played a massive role in devaluing small and medium sized websites (and the online ad business) and we’re certainly not going to be the ones getting any publishing deals. None of it sits well with me.

Google Unveils Music AI Sandbox

Google

Ty Pendlebury, writing for CNET:

Google has unveiled a new music-making tool it calls Music AI Sandbox, which enables loops to be created via AI prompts, as part of its I/O 2024 conference.

The tool, shown briefly today in a Google I/O video, appears to accept text input and provides short audio clips or “stems” based on the prompt, complete with a waveform representation of the invented sounds.

Spotify Recommending A.I. Generated Music

Spotify has been recommending “A.I. generated music” to some users:

My favorite example of this is AI music spreading across on Spotify right now. A user on X this week spotted an Artist page called Obscurest Vinyl that was promoted by Spotify’s Discovery Weekly.

The story behind the page is interesting. Obscurest Vinyl started as a Facebook page that would photoshop fake album covers for classic records that didn’t exist. The page recently shifted into posting AI songs to go with the fake album covers. As one commenter noted, you can tell the songs are AI because most of them feature bass and drum parts that don’t repeat in any discernible pattern. The account also regularly fights with users on Instagram who gripe about it using AI. 

Look, I think songs titled things like, “I Glued My Balls To My Butthole Again” are, honestly, pretty funny, AI or not. But they’re being uploaded to Apple Music and Spotify, which is where the snake starts to eat its own tail. Popular AI music generators like Suno clearly have datasets that include at least some copyrighted material (likely a lot). Which means, in this instance, Spotify is promoting and monetizing an account using an AI likely trained on the music that’s been uploaded to their platform that they don’t actually pay enough to support the creation of. And this is happening across every corner of the web right now.