”Bot Prownies” is a ten-track album of frantic skate punk tunes, complete with high speed drums, palm-muted guitars, and sneering vocals.
Though you could mistake much of it for a punk band’s sloppy demo tape, the record wasn’t created by human musicians. Instead, it was generated by a recursive neural network, a software framework that studies a set of input data — in this case the raw audio from the 1994 album ”Punk in Drublic,” by California skate punk group NOFX — then breaks it down into mathematical patterns and generates its own variation.
The algorithm was created by Dadabots, a pair of researchers named CJ Carr and Zack Zukowski who use machine learning to create music. Last year we covered about one of the duo’s earlier efforts, a record that successfully captured the screeching sound of black metal.
True to its genre, the black metal album was essentially noise, which helped mask its rough edges. In comparison, "Bot Prownies" is an unmistakable step forward. It captures much of NOFX’s melodic and stylistic sensibilities, and though the lyrics are gibberish, the rhythm and chord transitions are at once unique and weirdly evocative of its source material. During its best moments, it’s a glimpse of compelling synthetic creativity.
The caveat, as with many artistic projects created with the aid of artificial intelligence, is that there was a good deal of human curation. The Dadabots algorithm generated 900 minutes of audio, from which Carr and Zukowski selected the best bits — so the effort is more of a collaboration between humans and machines than a pure product of AI.