The following MBW Views review comes from Ed Newton-Rex (illustration), CEO of Fairly Training, an ethical generative artificial intelligence nonprofit.
Newton-Rex is a senior expert in the field of gen-AI, the former VP of Information at Stability AI, and the founder of JukeDeck (acquired by TikTok/ByteDance in 2019).
In this column, Newton-Rex argued that “riceUSICs produced using artificial intelligence products that are not licensed for training materials should be banned [from DSPs] Or should be given less weight in royalty calculations and recommendations…”
Leave it to Ed…
In April, when I wrote an article highlighting the striking similarities between Suno’s work and copyrighted music (and later when I did the same for Udio), I gave them the benefit of the doubt. It’s possible they signed a deal that allowed them to be trained in major label music. Theoretically, it’s even possible – though unlikely – that they had no training at all in the copyrighted music, and the countless similarities reach an uncanny level of coincidence.
But now, there is no room for doubt. The Recording Industry Association of America’s lawsuit against the two companies shows that there was no such training agreement. Both companies’ responses to the lawsuit acknowledged — both using the same language — that the recordings of their training “presumably include recordings for which the rights are owned by the companies.” [major record labels]”.
Suno’s response went a step further, saying their “training materials essentially consist of all music files of reasonable quality that are accessible on the open internet, subject to paywalls, password protection, etc.”
When it comes to generating artificial intelligence, streaming services will have to decide what content to allow on their platforms, one day. That time is now.
So far, Spotify has no policy explicitly banning AI-generated music. In 2023, Daniel Ek stated that tools that imitate artists are unacceptable; these may be prohibited by the company’s deceptive content policy (the wording is not entirely clear). However, in the same interview, Ek specifically pointed out that artificial intelligence music No Directly passing off the artist as something they wouldn’t ban at this stage.
There are signs that AI music is spreading across the platform as a result. Chris Stokel-Walker recently wrote an article for Fast Company about some bands with hundreds of thousands of monthly listeners that are suspected of being generated by artificial intelligence. Users of these artificial intelligence music platforms revealed that they are sharing artificial intelligence music to DSPs.
People are reporting that music recommendations in Discover Weekly playlists on Spotify are apparently generated by artificial intelligence. This month, an AI-generated song ranked No. 48 on the German pop charts and has been played more than 4 million times on Spotify to date.
For DSPs to continue to allow this behavior is to actively allow the exploitation of musicians’ copyrighted works without permission.
To quote the more than 200 artists who signed an open letter on artificial intelligence in music earlier this year: “Some of the biggest and most powerful companies are using our work to train artificial intelligence models without our permission. These efforts are The immediate purpose is to replace the work of human artists with a large number of “sounds” […] This significantly dilutes the royalty pool paid to artists. This would be disastrous for many professional musicians, artists, and songwriters who are just trying to make ends meet.
So far, people have wondered whether Udio and Suno are doing what these artists feared: training their music. This doubt has now disappeared.
These artists warn that the royalties paid to human musicians are being diluted when DSPs distribute music produced using artificial intelligence models trained on the work of unlicensed musicians.
Musicians’ royalties are being diluted by products made using their work against their will. DSP is facilitating this.
What can be done?
First of all, it’s worth mentioning that I don’t think DSP should ban all AI music. There are clearly good use cases for artificial intelligence in music creation; and if the training material is licensed, those use cases are worth supporting, at least in my book. (I do think there will be music streaming services Do A clear rejection of all artificial intelligence music, as Kara does in Image Space. It will probably do well. But most DSPs don’t take this blanket approach, and for good reason.
As a bet, DSPs should follow the example of other media platforms such as Instagram and TikTok and label AI-generated content.
This way, music fans can at least choose what they listen to, and therefore what they support. Uploaders are asked to label the AI music they upload and introduce a post-upload review process for missing tracks. This is totally doable. You hope that most uploaders are honest – in general, people tend to prefer honesty – and for those who aren’t, there are a number of third-party systems that can detect AI music with high accuracy.
Of course, there is also the question of how much involvement of artificial intelligence should be required to trigger the application of tags.
Typing text cues and distributing the output on Spotify is obviously very different from using a MIDI generator for inspiration.
But this difficulty is not insurmountable, nor is it a reason to avoid labeling altogether. DSPs just need to be clear about their policies and apply them equally to everyone. As a starting point, tags can be applied if any generative AI was used in the creation of the track.
But I think DSP should go further than labels. Music produced using artificial intelligence products that do not authorize its training materials should be banned or have less weight in royalty calculations and recommendations.
Otherwise, it’s going head-to-head with the music it’s trained on – and that’s unfair. (If at this point you’re tempted to say, “But humans can learn from and compete with existing music” – please don’t. Training an AI model is completely different from human learning, and it makes a huge difference to the market.)
“DSP should go further than labels. Music produced using AI products that don’t license its training material should be banned or have less weight in royalty calculations and recommendations. Otherwise, it will go head-to-head with the music it was trained on — -That’s not fair.
One problem here is that we don’t have a detailed list of which AI products fall into this category because there is currently no requirement for AI companies to reveal what they are trained on. (There should be, but there isn’t.)
Udio and Suno acknowledged this in court documents, but there may be other companies taking the same approach. However, this is no excuse for not taking action at all. DSPs should do their own due diligence, and if the balance of probabilities is that an AI model was trained on unlicensed music, I think it’s fair to subject music produced using that model to different rules.
Some would say that DSP should wait until these lawsuits make their way through the courts before deciding how to act.
But royalties are now being diluted. There is ample precedent for DSPs to enforce content policies on principle rather than because of specific legal rulings. For example, according to Spotify, it “invests heavily in detecting, preventing and eliminating the royalty impact of artificial streaming” (the idea that people play tracks on repeat throughout the night to get more plays) and taking actions to reduce the royalty impact. “Bad actors” use white noise recordings to manipulate the system.
The company believes such changes “could generate approximately $1 billion in additional revenue for emerging and professional artists over the next five years.”
If this is the purpose, why not take action against music produced using artificial intelligence models trained on the work of unlicensed artists? Like white noise, it is used to trick the system and redirect royalties. Unlike white noise, it was created using the work of competing artists.
I agree with Daniel Ek that there is a controversial middle ground when it comes to regulating AI music. I very much don’t want to ban all AI music: when it’s permission-based, there are definitely some use cases that are a net positive for musicians.
But if DSP’s mission is to “give one million creative artists the opportunity to make a living from their art,” I think it’s obvious that they should draw the line at diluting by recommending music made with products that exploit the work of other musicians without their permission Royalty pool in the process.
The DSP may delay deciding how to respond to this new threat to musicians until they are forced to make a decision. But if they don’t act soon, I suspect it won’t be long before we see the first artists pulling their music from these platforms in protest.global music business
