Spotify's Venture into AI
Spotify's Venture into AI-Generated Playlists with Prompts
Table of Content
After introducing AI-translated podcasts and an AI-powered DJ feature, Spotify is now exploring AI-generated playlists. Recent discoveries within the app's code, brought to light by tech enthusiast Chris Messina, hint at the possibility of users creating generative AI playlists using prompts. This post explores these intriguing developments and their potential implications for music lovers.
Spotify, the popular music streaming service, is reportedly working on a new feature that would allow users to create playlists using artificial intelligence (AI). The feature, which is still in development, was revealed by Chris Messina, a tech investor who found references to “AI playlists” and “playlists based on your prompts” in the app’s code.
How Would AI Playlists Work?
According to Messina, the AI playlists feature could be related to the Blend genre, which lets users mix their musical tastes with others to generate a playlist that everyone likes. Users could also invite others to create AI playlists together, based on the code references.
The feature would use generative AI to create playlists based on user-provided prompts, such as a genre, like Cottagecore Indie Mix, Bubblegum Pop Mix, Discofox Mix, Feel Good Driving Mix, Fun Road Trip Mix, Travel Mashup Mix, and others that match the description.
Spotify’s History with AI
However, not all of Spotify’s AI features have been made public. Spotify also has a feature called Niche mixes, which lets users create playlists based on descriptions, but the company said that these are not AI-powered, but rather based on personalization algorithms.
Spotify also has a large team of researchers and engineers working on various aspects of AI, such as language models, generative voice, and personalization. Spotify’s head of Personalization, Ziad Sultan, said that the company is constantly exploring “all the possibilities across AI” to improve its product offering and offer value to users.
Spotify declined to comment on the AI playlists feature, saying that it does not comment on speculation around possible new features. However, the discovery of the code references suggests that Spotify is serious about using AI to create more engaging and personalized playlists for its users.
Interestingly, Spotify may have been laying the groundwork for AI playlists created with prompts through a feature called "Niche mixes." This feature currently allows users to construct unique playlists based solely on descriptions.
These Niche mixes permit users to specify various parameters, from genre and vibe to aesthetics like "Cottagecore Indie Mix" or "Fun Road Trip Mix." However, it's worth noting that, despite initial appearances, Spotify clarified that Niche mixes are not AI-powered but are instead driven by the company's personalization technology and algorithms.
Chris Messina's Findings
Chris Messina's findings suggest that the new AI playlists may also be generated using prompts. However, as of now, this feature has not been publicly unveiled and can only be inferred from the code.
Messina speculates that this feature might be linked to Blend, as code references indicate the possibility of users inviting others to collaborate on creating AI playlists.
All the lines of code hinting at AI playlists were discovered in the latest build of the Spotify app, indicating that this is a new, in-development feature. While not all internally tested features make their way to the public, this development underscores Spotify's commitment to exploring the role of AI in music personalization.
The AI Vision at Spotify
Spotify has previously hinted at its ambitious AI roadmap, suggesting that the AI DJ was just the beginning. The company has an extensive team dedicated to personalization and machine learning, including a substantial research team exploring the possibilities of Large Language Models and generative voice.
Spotify's head of Personalization, Ziad Sultan, emphasized their commitment to understanding "all the possibilities across Large Language Models, across generative voice, across personalization."