Artificially Intelligent Band Recommender Takes Your Tastes Into Account
The music-recommending website Sage uses artificial intelligence to review one's favorite band and most hated band and give suggestions that match the underground acts closest to those tastes.
But working with the recommender can also elicit suggestions based on more subtle shades of music partiality — it's able to group artists together on each side of the input and answer with the rock, punk and metal acts that line up. Still, taking one band each from the most extreme ends of one's likes and seeing the results is perhaps the most singularly fascinating part.
It's all the project of videographer and music curator Sunny Singh, the documentarian mastermind behind the punk concert video archive hate5six. Interestingly, he just happened to be a data scientist before he became the punk scene's premier archivist. But does Sage really work? I wanted to find out.
For an example, I punched in that I like At the Drive-In but don't like Slayer. That doesn't represent my actual tastes — I just used those to test the system. The results returned disbanded D.C. post-hardcore act Q and Not U as my top suggestion with a 63.1 percent match, followed by Lungfish (56.3 percent), Cheap Girls (53.5 percent) and Bear vs. Shark (52.5 percent).
How does it do that? According to a very technical blog from 2017, back when Sage first launched, the recommender analyzes data called "graph embeddings" (the name literally stands for "Sage Analyzes Graph Embeddings") through listening vectors it scans to "clearly define a taste profile and receive suggestions in a principled manner." For the more mathematically inclined, feel free to read more about it here.
"We've been able to leverage publicly available data about communal listening habits across over 200,000 bands and developed a novel model for finding new music," Singh explained. "The model has been able to learn fairly robust mathematical representations of bands that preserves their 'context': bands that share members, have similar tempos, are lyrically and thematically related, tend to cluster together in the embedded space. This enables the user to define taste profiles capturing what they do and don't like, and that corresponds to a well-defined set of mathematical operations on the embedded representations of bands."
Ready to put it to the test? Try out Sage here.