Do androids dream of electric betrayal? That's just one question looming over this cover of "Jolene," made by the musician Holly Herndon using her "deepfake" digital twin Holly+, built to replicate the artist's own singing voice using machine learning technology.
"Jolene" is a song that's almost primal in its insecurity, and the high, fluttering way in which Dolly Parton delivers her pleas so heartbreakingly desperate. But there's a freaky, almost too well-roundedness to Holly+'s interpretation. It's as if every edge of what was once a vulnerable, living voice has been sanded down into as smooth a surface as possible, like a glinting granite countertop lining the pristine kitchen of a suburban McMansion. Listen closely and the sound of a deep, simulated breath emerges like a vapor in between a few verses, always for just a second too long.
So, Holly+'s "Jolene" is a little creepy. But the possibilities of Holly+ are worth paying attention to, even if the music "she" creates here — or computes — sounds dull. Holly+ joins a growing and often misunderstood landscape of experiments in music and artificial intelligence, many of which tend to overly fetishize mimicry and duplication of other artists' works and likenesses (some living, some dead). What's compelling about Herndon's project is her centering of herself, and the experiments in consent and ownership she's designed in a field where the legality and boundaries of both of those things are still disturbingly in flux, even as celebrities and politicians have their likenesses continually manipulated with frightening accuracy.
Letting listeners inhabit her vocal likeness and make music through her on her own terms, Herndon demonstrates what control for artists can look like when it comes to ever-evolving deepfake technology. And what better way to communicate the perils of taking something that isn't yours than through the words of "Jolene?" If only Holly+, masterful little replicator that she is, could communicate the emotional weight of what that feels like. Maybe one day, with some fine-tuning, she will.
Copyright 2022 NPR. To see more, visit https://www.npr.org.