Apple’s AI Playlist Playground is bad at music
Apple Music's new AI playlist generator struggles with basic prompts, delivering inappropriate and inaccurate song selections.
Apple's new AI-powered Playlist Playground feature for Apple Music is currently in a rough beta state, struggling with fundamental aspects of music curation. In tests by The Verge, the tool failed to accurately interpret prompts for specific genres like 'atmospheric instrumental black metal,' instead serving a chaotic mix of vocal metal, ambient electronics, and doom jazz. It also demonstrated a poor grasp of geography, suggesting a band from South Dakota for a playlist requesting music from the American South. The AI's understanding of time periods was equally flawed, labeling songs from 1998 as 'modern' hip-hop.
Perhaps most concerning was its failure to filter lyrical content. A prompt for 'kid-friendly modern hip hop' returned censored versions of explicit songs, including one with graphic sexual references. Even for broader prompts like 'industrial-influenced dance punk,' the results were off-target, favoring classic industrial acts over the requested modern sound. While still in beta, Playlist Playground's current performance suggests it lacks the nuanced understanding of music metadata, context, and user intent required for a functional recommendation engine, lagging significantly behind existing AI tools from competitors.
- Fails on genre prompts, mixing black metal with ambient and electronic tracks.
- Demonstrates poor grasp of geography and time, mislabeling locations and decades.
- Inappropriately includes explicit lyrics in 'kid-friendly' playlists, showing a lack of content filtering.
Why It Matters
Highlights the significant gap between AI hype and practical utility in creative domains, even for a tech giant like Apple.