97% Struggle to Identify AI Music, but It’s Less Problematic Than Feared

Deezer recently partnered with Ipsos to conduct an intriguing study on music perception. The experiment revealed that a staggering 97% of participants struggled to differentiate between fully AI-generated music and that created by humans. However, this alarming statistic requires further context.

Study Overview

In this experiment, 9,000 individuals listened to three music tracks and were tasked with identifying which, if any, were entirely AI-generated. If a participant failed to identify all three correctly, they were classified as unable to discern the difference. This means that even if they identified two tracks correctly, their overall ability to distinguish was deemed inadequate.

Independent Experiment

To further investigate the claims, I conducted a secondary, informal study. I gathered ten participants and presented them with the same three tracks. From this group, only one person accurately identified all three tracks. However, aggregating the individual responses, participants successfully recognized AI-generated music about 43% of the time.

Participant Reactions

Many participants were taken aback by their inability to distinguish between AI and human music. Survey results indicated that 71% felt surprised by their performance. Furthermore, 51% expressed discomfort at the indistinguishability of AI-generated art from human creations.

Concerns About AI Music

• 51% of respondents worry AI will produce more “low-quality, generic” music.
• 40% stated they would skip AI music without listening.
• 80% desire clear labeling for AI-generated content.

Deezer’s Transparency Measures

Deezer’s response includes developing a system for automatic detection and labeling of AI-generated music. This system targets popular AI models like Suno and Udio. Moreover, tracks classified as AI are excluded from algorithmic recommendations on the platform.

Industry Responses

In contrast, Spotify has announced measures against AI impersonation and plans for a standardized credits system but has not committed to explicit labeling of AI content. They advocate for a nuanced approach to AI transparency, which relies on honest disclosure from artists and labels.

Volume of AI-Generated Music

The amount of AI-generated music is rapidly increasing, with Deezer reporting over 50,000 tracks per day. This figure comprises more than 34% of the new music added to the platform. However, these tracks only represent 0.5% of overall streams, with most being considered fraudulent.

Future Implications

Despite concerns, experts like Manuel Moussallam from Deezer believe there is no imminent threat to human creativity in music. He asserts that while AI content is on the rise, it does not drastically alter the listening experience for users. Music created by humans will continue to resonate with audiences.

In summary, while the struggle to identify AI music appears significant, the reality is nuanced. The integration of AI into music creation raises ethical and transparency issues, necessitating collaboration among creators, labels, and streaming platforms.