Spotify shares its secret sauce: 3 layers of AI-driven algorithms

Spotify’s AI-driven algorithms have been praised for their accuracy and feared for potential bias. With 500 million monthly active his users, the workings of its algorithms have had a profound impact on music and global culture.

Spotify now details how it uses machine learning to create experiences for each user on its platform, including what’s playing and why the new WSJ video (see below). .

Spotify’s secret sauce lies in how it layers one algorithm on top of another to discover incredibly accurate and personalized music recommendations.

Layer 1: Collaborative Filtering

Collaborative filtering looks at when a song is played or a playlist is created next to another song. “Think of it like making a map.”

Layer 2: Content-based filtering

Content-based filtering collects metadata and performs basic audio analysis to determine danceability, loudness, etc. to describe the sonic characteristics of the track. We also analyze the lyrics and adjectives used to describe the track in online articles and logs.

One of the problems with content-based filtering is that it can reinforce existing musical and cultural biases in a kind of feedback loop.

Another is the “cold start problem”. In other words, there is a lack of data for new artists. Here, Spotify’s human curators step in to find the best new music.

Layer 3: Reinforcement Learning

Reinforcement learning is the latest layer and is gaining importance. This allows Spotify’s entire recommendation engine to learn and improve based on your feedback.

Tomorrow: Spotify’s Algorithm From an Artist’s Perspective

This 7-minute video is an intensive study of Spotify’s algorithm.

• 0:00 Spotify is known for personalized playlists created with a recommendation algorithm
• 0:37 Evolution of Spotify’s Recommendation Algorithm
• 1:42 Spotify uses user data to create a map of songs and artists
• 2:59 How Spotify’s content-based filtering technology works
• 4:09 The Danger of Algorithmic Bias
• 6:19 Spotify’s future plans in AI

Bruce Horton Founder and Editor of Hypebot and MusicThinkTank, Senior Advisor to Bandsintown, President of Skyline Artists Agency, and Professor at Berklee College of Music.