Global streaming service Deezer partnered with Ipsos to conduct a consumer survey to discover popular perceptions and attitudes toward AI-generated music. Nine thousand people in eight countries were questioned. The topline takeaway: “It revealed a clear desire for tagging 100% AI-generated music and making sure artists and songwriters are being fairly treated and paid if their music is used to train AI-models.”
Participants listened to three tracks and flag any that sounded “fully AI-generated.” No matter what presumption a person might bring to the expected outcome, the result was startling: 97% of respondents failed to correctly sort all three.
In a secondary question, 71% were surprised by their results, and 52% “felt uncomfortable by not being able to tell the difference.” (Details HERE.)
Worried
“I’m starting to get a little worried.” That’s concert violinist Ray Chen, who put his ear to the test of identifying AI classical music when compared with acoustic recordings. (HERE) Chen listens to clips music from the Baroque, Classical, Romantic, and Contemporary period, and film scores. (You can play along with him.)
Chen’s results are mixed — a conclusion which verifies how skillful AI music has become, compared to human performance. “I’m starting to get a little worried,” he says. Readers who keep a finger on the pause button can play along. (This demonstration is a year old, but appeared on our screen today.)