FrontierThe story, in brief

MuseNet

OpenAI's MuseNet generates 4-minute symphonies across 10 instruments using the same transformer tech as GPT-2.

Illustration of a transparent lens revealing connected networks across layers of paper.
Exploring the next frontier of AI research.AI illustration by KeyNews
The KeyNews take

Why it matters

OpenAI demonstrates that large-scale transformer models trained on sequence prediction can generalize beyond text to music composition, validating the universality of the GPT-2 architecture and expanding the frontier of what unsupervised learning can accomplish.

The key facts

6 to know
  1. MuseNet generates 4-minute compositions

  2. Supports 10 different instruments

  3. Trained on hundreds of thousands of MIDI files

  4. Uses same transformer architecture as GPT-2

  5. Learns through next-token prediction (unsupervised learning)

  6. Combines styles: country, Mozart, Beatles

Go to the source

OpenAI Blogopenai.com

Publisher excerpt: We’ve created MuseNet, a deep neural network that can generate 4-minute musical compositions with 10 different instruments, and can combine styles from country to Mozart to the Beatles. MuseNet was not explicitly programmed with our understanding of music, but instead discovered patterns of…
Read original report
Back to today's editionMore frontier news

Keep reading

Related stories

More from Frontier