Human-AI Collaboration for Creativity with Devi Parikh - #399

Dance Discovery

Devi Parikh explores the innovative use of AI in discovering dance movements that synchronize with music without relying on pre-existing dance data. The project aims to generate movements that align with music by analyzing its structure and creating a sequence of movements that match similar points in time 1. Devi emphasizes the importance of visualizations in making these movements more inspiring and appealing 2.

Our interest in this was not to figure out how we can get a humanoid to stay stable and learn the laws of gravity or anything of that sort... We were interested in this question of if we produce movements that are just in sync with the music, that's all the constraint is. What does that look like? Does that look interesting or not?

--- Devi Parikh

This approach allows for the emergence of unique dance forms that are not constrained by existing data, offering fresh perspectives on movement and creativity.

   

Music Analysis

The analysis of music's structure plays a crucial role in creating synchronized movements using AI technologies. Devi Parikh describes a method that uses autocorrelation in an acoustic feature space to align music and movement, ensuring that the movement's autocorrelation matrix mirrors that of the music 3. This technique allows for a variety of visualizations, from simple stick figures to complex geometric patterns, all driven by the same underlying model 4.

It's the same thing that we're looking at in the movement that we also have a similar autocorrelation like matrix for the movement. And we are trying to say that the autocorrelation matrix of the movement should be similar to the autocorrelation matrix of the music.

--- Devi Parikh

Such flexibility in visualization not only enhances the aesthetic appeal but also inspires new forms of artistic expression.