Photo illustration: Intelligencer. Photo: YouTube
Earlier this month, Spotify announced a new deal with Universal Music Group, the largest of the three major companies, that will add the ability for “fans to create covers and remixes of their favorite songs by participating artists and songwriters.” The tool, “powered by generative AI,” promises “unlocking additional revenue streams” and “new ways to drive discovery.” The companies say users will have to pay a fee for access, and a portion of that money will be donated to participating artists. The feature will allow Spotify to “turn one song into 10,000 songs,” according to the company’s co-CEO.
What this means for Spotify is clear. Streaming giants can offer a legally sanctioned sharing economy that is only accessible on their services. If instant remixing of popular music becomes the next big thing, there will be more listeners, more subscription revenue, and it will be harder to escape the platform. And if this I don’t It turns out that’s what people want, but, well, no big deal.
But what about record labels? In its announcement, UMG suggested that this is just one example of “leading the industry through technological change” and that the deal is an example of “responsible AI” in a time that many of the label’s artists are concerned about. In an interview with financial timesSpotify co-CEO Alex Norstrom was not so subtle. “There are a lot of fraudulent attempts at this,” he says. He went on to say that Spotify is both “legal” and “controlled,” as opposed to a flood of “sloppy” AI-generated music. (Although unstated, much of it is flowing into Spotify itself.) Record labels and the artists they represent are acting in response to the threat. The idea is that people will do this anyway, so why shouldn’t Spotify try to formalize it?
This isn’t too far removed from Spotify’s original concept of providing a legal streaming framework for record labels concerned about copyright infringement and online music, but it comes with similar corollaries. But in the late 2000s, Spotify was just one of many startups pitching into a legacy industry scared of file sharing. Today, Spotify is a technology giant and a dominant platform with substantial influence over the companies and artists it is forced to partner with.
In other words, it’s similar to YouTube. And wouldn’t you know it, an even more ubiquitous platform released its own take on AI remixing, which would happen anyway, around the same time as Spotify. “Gemini Omni is an exciting upgrade that makes it easier to participate in trends and conversations on YouTube,” the company said. “Remix targeted short videos by adding your own prompts and images to create entirely new visions, such as changing the scene to a ’90s vibe or inserting yourself alongside your favorite creators, while keeping the context of the original video intact.” It also announced that it will be expanding its feature “Likeness Detection,” which lets you track and control the use of your face in other people’s video posts.
YouTube, like Spotify, has a lot to gain and little to lose. If users see this as a gimmick, like OpenAI’s Sora, the company can move on.
Photo: YouTube
However, unlike in the case of Spotify, the majority of YouTube’s content was created expressly for YouTube and uploaded by its creators. Both Spotify and YouTube’s AI features are opt-in for artists and creators. In addition to constituting an AI experiment with a large user base, AI Remix gives the Google-owned platform the opportunity to further build its own bizarre intellectual property and rights regime that extends beyond personal creations to identities. In part, this is the purpose of similarity detection.
As AI-generated content continues to evolve, we’re working to build an environment where people can upload content to YouTube while managing their similarity… Today, we’re excited to announce that in the coming weeks, we’ll be expanding similarity detection to all eligible creators over the age of 18.
There is precedent for this in corporate history as well. In the late 2000s, as Google faced a flurry of lawsuits from TV and music companies that had watched YouTube grow as users uploaded countless show clips and music videos, it began working on an automatic detection system that would become known as Content ID. While Content ID is designed to help watermark and distribute royalties to individual content (songs and movies, for example), similarity detection with its biometric onboarding process is like YouTube’s copyright for the entire user, storing eye-to-nose-to-mouth measurements in a database (to simplify things considerably), just like music tracks.
For now, similarity detection is presented as an AI solution to an aggravating and relatively specific AI problem: fake unauthorized videos. Industry insiders say the appeal of AI Remix is that it is clear and intuitive. Of course, some people may want to adapt the songs to their own tastes or insert themselves into the story. — I suspect that if these tools were introduced everywhere, the results would be less obvious and stranger than their creators hope. But features like this also add something new to the vast range of material people commit to Google, which for many users is the Internet-wide identity service. They already entrust their emails, messages, documents, photos, and search history to this platform. Now, in a world where AI remixes are endless and inevitable, Google says AI can also help you manage your face.
