97% of people have difficulty recognizing AI music, but it’s not as bad as it seems

Streaming service Deezer recently conducted a trial with the help of research firm Ipsos. The result – that 97 percent of people The indistinguishability between fully AI-generated music and human-made music was alarming. But it’s also not the whole story.

In the survey, 9,000 participants listened to three tracks and were asked to guess which, if any, of them were entirely AI-generated. If the participant fails to guess all three correctly, they are placed in the failure pile. This means that if you get two out of three right, Deezer and Ipsos still say you can’t tell the difference between completely AI-generated music and real music.

Deezer sent me the three tracks they used in the study, so I decided to conduct my own (less scientific) experiment. I had 10 people listening to the same audio clips and I gave them the same prompt. People have had a hard time identifying songs that are entirely AI-based. Only one person got all three correctly. But if I had not aggregated the answers, the results would have been much less dire. People were able to successfully identify whether a track was AI-related or human-generated with 43% success.

It’s also worth noting that several people told me that one of the songs was so terrible, and so obviously AI, that they thought it was a trap and guessed it was real.

Not surprisingly, participants in the Deezer study were a little surprised at how poorly they performed. 71% were surprised by the results, and 51% said that their inability to distinguish between AI-made art and human-made art made them feel uncomfortable.

Opinions were divided on the impact, with 51% believing that AI will lead to the creation of lower quality, more generic music. Somewhat shockingly, only 40% said they would skip AI music without listening to it if they encountered it intentionally.

However, one area that most people agreed on was the need for transparency. Eighty percent want AI-generated music to be clearly labeled. For now, this has been the approach taken by Deezer. He created a system that can do this automatically Discover and label 100 percent AI-generated content The most popular models are Suno and Udio. Deezer also excludes music that has been labeled as AI from its algorithmic recommendations.

Spotify recently Announce steps to Combating AI decline on its platform, but stopped short of saying it would explicitly label AI content. It has announced policies around AI impersonation and a new spam filter that will keep many of the worst actors off its platform. But rather than a blanket rating, she is working to create a standardized credit system, saying: “The industry needs a nuanced approach to AI transparency, so that you don’t have to label every song as either ‘AI’ or ‘non-AI.’” However, this system will rely almost entirely on labels and artists openly disclosing when songs use AI, even if it is simply to aid in the mixing process.

Manuel Muslim, Research Director at Deezer, says: Edge There is a small gray area around hybrid content that might use AI elements. But he says this is “not a technical issue. It’s a transparency issue and an ethical issue” that will require all parties involved, from creators to music distribution services like DistroKid to streaming platforms, to act responsibly.

What is clear is that the amount of music being uploaded by AI is staggering, and it is growing. Deezer says it receives more than 50,000 AI-generated tracks daily, representing more than 34 percent of the music added to the service.

Muslim says that although the company receives a staggering amount of AI content, it only represents 0.5 percent of streams, and the vast majority of that small slice is fraudulent. While he admits it poses some challenges for Deezer due to its size, he doesn’t think it fundamentally changes the experience for users. “Human beings will continue to create music, and they will continue to listen to music made by real artists,” he says.

Holly Herndonwhich was used Custom AI models She agrees on this widely in her own music Edge that “just because anyone can create polished stuff doesn’t mean anyone will care about those songs. Artistic practice is much more complex than that.”

Seventy percent of survey respondents believe that fully AI songs pose a threat to musicians’ livelihoods, and 64 percent believe that AI can lead to decreased creativity. But Musallam was less pessimistic, saying: “We are not heading towards a future in which humans are removed from the creative process, but rather artificial intelligence will be integrated into the creative processes.”