In an industry known for disruption, it should come as no surprise that music streaming services are pushing the boundaries of technology to create exciting possibilities for listeners.
Currently, the hottest area of development is to revolutionize music discovery and personalization. All of the top players in streaming are working with artificial intelligence and deep machine learning in bids to build better, more intuitive playlist algorithms.
Spotify’s approach, called collaborative filtering, is to collect as much data as possible on a user’s listening behavior and then compare it to that collected from other users. To augment this data and improve recommendations, the streaming service acquired Echo Nest, the leader in machine learning and music discovery, which gathers intelligence about new music posted to blogs, news websites, and social media.
Spotify’s highly popular Discover Weekly feature, for example, suggests new music based on a user’s listening habits. The service said 40 million unique users streamed more than five billion tracks in the 10 months after the feature’s launch, and it recently launched a follow-up to this feature, called Daily Mix. Like Discover Weekly, it mines a user’s listening data to create a personalized playlist designed to present hours of music that a user will love.
Pandora takes a slightly different approach, deploying a combination of human expertise and algorithms. Music experts tag songs with hundreds of characteristics, from genre to tonality. The company’s algorithm then matches these characteristics so that when a user selects a particular artist or song, a “radio station” of music with similar attributes is created.
By contrast, Apple emphasizes playlists curated by human experts, including musical artists themselves. While it isn’t using artificial intelligence, relying on human tastemakers is an approach that differentiates the service.
Still, to gain a competitive edge, the services are taking artificial intelligence a step further by investing in “deep machine learning.” This technology offers greater promise for classifying music and making more accurate recommendations to users. For example, Spotify is working with the technology to analyze the song itself, rather than simply its metadata. Google and Pandora also have hired deep learning experts and are working with “artificial neural networks” to create better playlists.
Beyond music playlists, AI and machine learning are paving the way for other types of music applications. For example, it’s now possible for computers to write songs. Scientists at Sony’s CSL Research Laboratory released the very first pop song composed by an artificial intelligence system called Flow Machines. Drawing from an enormous database of songs varying in styles, the system can choose minuscule elements from a range of tracks to compose a unique song.
This technology opens the door for brands, small businesses and individuals looking to augment advertising, videos and other marketing pieces with original but inexpensive music. Jukedeck, a London-based startup, offers unique, royalty-free compositions designed by algorithms and written in just seconds.
Another innovator, Brain.FM, is focusing on how music can be used to help individuals with ADD, anxiety and insomnia. Working with neuroscientists, they’ve created AI software for audio brainwave training, designed to improve focus, relaxation and sleep.
While music lovers may be the first to benefit from such innovations, it’s clear the combination of music and AI will offer many possibilities beyond personalized playlists – potentially disrupting other industries beyond music.