OpenAI Is Joining Amazon in a nascent group of major tech companies to invest astronomical sums of money in machine learning projects for generating music.
This example output from OpenAI’s new Jukebox demo probably illustrates it best:
OpenAI situates this track as the output of their waveform generating neural network, which has been conditioned on music by Kanye West.
The result is actually a very garbled hip hop audio with clearly audible excerpts from Eminem’s Lose Yourself. It is more revealing of Deep Neural Networks’ tendency to obscure and appropriate the labor used in their creation than it is impressive that they made Eminem sound a little bit like Kanye West.
As was the case for Amazon’s DeepComposer, George Lewis’ writing about music tech is probably the best lens through which to study OpenAI’s Jukebox. Lewis published a paper in the Leonardo music journal in 2000, which includes this quote:
“As notions about the nature and function of music become embedded in the structure of software-based musical systems and compositions, interactions with these systems tend to reveal characteristics of the community of thought and culture that produced them.”
We should understand the meticulously researched, astronomically expensive, machine learning technology on display here as representative of the “community of thought” at OpenAI.