Spotify’s Speech Monitoring Tech Sounds Creepy as Hell



Due to the success of curated playlists like its Uncover Weekly characteristic, Spotify is at all times searching for methods to enhance its algorithm. However as Pitchfork reports, its newest transfer in the direction of that finish goes above and past: the streaming service has been granted a patent for expertise that can use recordings of customers’ speech and background noise to find out what sort of music to advocate them.

Spotify filed for the patent again in 2018, but it surely was authorised on Jan. 12, 2021. The submitting outlines some potential makes use of of the expertise, together with detecting “intonation, stress, rhythm, and the likes of models of speech” from a consumer’s voice as nicely as having the ability to determine their emotional state, gender, accent, age and atmosphere (that’s the place the background noise comes into play) based mostly on the audio recording to assist curate music for them.

“What is required is a wholly totally different method to accumulating style attributes of a consumer, significantly one that’s rooted in expertise in order that the above-described human exercise (e.g., requiring a consumer to supply enter) is not less than partially eradicated and carried out extra effectively,” the patent submitting reads.

After all, along with elevating a slew of privateness considerations, this begs the query: are you able to really inform what sort of music somebody desires to listen to based mostly on their voice? It appears as although any curation that could possibly be completed based mostly on gender or accent, for instance, could be merely rooted in dangerous stereotypes. (Are ladies going to be fed extra feminine artists than males are? Is somebody with a Southern accent solely going to get nation songs of their curated playlists?) Even one thing like emotional state looks like a precarious factor to attempt to base a musical suggestion round; some folks wish to wallow and hearken to unhappy songs after they’re feeling down, whereas others desire to choose themselves up with some upbeat tunes. How can an algorithm presumably know what somebody’s private preferences about that type of factor are?

We’ll have to attend and discover out. As Pitchfork notes, “It’s at present unclear whether or not Spotify has established a roadmap for the implementation of this tech into its desktop or cell apps, or what type this implementation may take. It’s also unclear if the expertise at present exists or if the patent is speculative.”

Subscribe here for our free each day publication.





Source link