Welcome to CO/AI

We slow AI down so you can speed up

Our mission is to make 1,000,000 people AI Literate

Today’s Letter:

OpenAI’s Music Generation Tool (4 min)

Today we break down early research from OpenAI: Jukebox, the neural net that generates music, including rudimentary singing, as raw audio

The Big Picture:

• OpenAI Released Jukebox: A model trained on 1.2 million songs able to create music in the style of popular artists across many genres

• Technological Marvel: Jukebox stands out for its ability to model music as raw audio, embracing a vast array of genres and artist styles, and showcasing unprecedented local musical coherence.

• Quality and Complexity: The model’s technique of compressing audio into a lower-dimensional space before generating music sets a new standard for audio quality and musical complexity in AI-generated music.

Jukebox Sample:

OpenAI has released several generated songs on SoundCloud as early as 4 years ago:

Why It Matters

• Creative Democratization: Jukebox democratizes music creation, offering tools to produce high-quality music, potentially transforming the music landscape by making production more accessible.

• Innovation and Collaboration: The platform opens up novel avenues for collaboration between human musicians and AI, suggesting a future where music is co-created with machines.

What do You think?

Thinking Critically:

• Generated samples still lack larger musical structures, suffer from noise artifacts, and render slowly

• Concerns about AI detracting from authenticity and emotional resonance of human-created art

• Worries about AI mimicking artists' styles without permission or attribution

Read the Full Study Here: