The audio-hacking duo, CJ Carr and Zack Zukowski, started using AI to create never-ending music streams on YouTube. The hackers launched this as a project named “Dadabots” and use a recurrent neural network (RNN) to create live streams on never-ending music. So, Dadabots’ recent live stream is an “infinite bass solo” which they created using two hours of bass improvisations by popular YouTuber and bassist, Adam Neely. You can check out the live stream right below.
How Does it Work?
Now, for the unaware, a recurrent neural network is a kind of a machine learning model that identifies music patterns, predicts the most common elements, and then reproduces them in its own way. So, Carr and Zukowski trained this model on bass materials by Neely to start a never-ending live stream of unlimited bass solos. The duo, following a few initial trials, limited the training dataset to fast-paced bass playing. This is because the RNN is more efficient in identifying faster tempos. As a result, the quality of the live stream improved by many folds and, according to the Dadabots team, this helped the AI to generate more exciting bass playing. Carr and Zukowski also note that sometimes, in the live stream, it may seem that the AI is playing two bass guitars simultaneously. This happens because of the high-temperature values of the bass notes. However, despite its imperfections, the AI does a pretty good job in reproducing the bass licks. So, if you are into fast-paced bass playing and want to use it in your projects, you can sample the audio from the live stream under a CC-BY license.
