One of our approaches to mimicking human activities through AI is the ability of systems to simulate our movements, the physical movements of our bodies.

We spend a lot of time talking about cognitive imitation—the way neural networks start to think like us. But what about all the physical activities we perform as humans?

At MIT, we are working on several projects applying AI to a specific type of activity – dancing.

Researchers such as Pat Pataranutaporn will speak about the “Dancing with AI” project, in which teams hope to develop “choreographic intelligence” using traditional dance practices from around the world. (Pataranutaporn is from Thailand and works on cyborg technologies.)

Quotes from Pat:

“(Staff have) been thinking about how we can break down these traditional dances into events that we can then teach to AI, so that AI can help us develop new choreographies and dance together with a human… it’s like bringing the tradition to life by creating a new cultural heritage.”

“At the moment we have a lot of models that can interpret text or images, but here we’re trying to develop a model that can actually interpret body points and be able to intervene and develop new choreographies. We find this really interesting because it captures what we call ‘choreographic intelligence’, the ability of the human body to respond to dance movements on stage. And it’s also really important that we try to acknowledge the non-Western cultural knowledge that can also be embedded in AI systems.”

This project is great, but MIT is doing even more.

This curriculum project is also credited to Brian Jordan, Jenna Hong, and Nisha Devasia, along with instructors Cynthia Breazeal, Randi Williams, and Hal Abelson, in a course titled “Democratizing AI for K-12 Education.”

And we have a project for MIT App Inventor that a team member describes as follows:

“Our amazing friends at YR Media published a fascinating interactive story about AI called ‘Can AI be taught to dance?’ that got us thinking. Is there an algorithm to describe, identify or measure a great dance? Can dance steps be quantified and measured? Can AI be used to help you improve your dance skills? In this project, you will learn how to use the new AI technology PoseNet to track key points on your body to create a skeletal model and develop some basic methods to quantify, measure and identify some dance steps.”

So if you’re the type of person who likes to boogie or two-step, or if you’re interested in the history of ethnic dances, you might find yourself consulting an AI unit to figure out more about your choreography.

As with so many of these technologies, many of them are partially trained by experienced choreographers for a human-in-the-loop approach…

If you want to know what’s going on outside of MIT, here’s an interesting article from Dance Magazine from last year where the author talks about ChatGPT applications, etc. Lauren Wingenroth responds to claims that dance-related AI will be helpful:

“If it sounds like the robot is protesting too much, that might be its attempt to acknowledge a growing existential concern as the dance world reckons with the current and potential future impacts of ever-increasing artificial intelligence. These technologies further complicate the dance world’s already frayed relationships with copyright, attribution, compensation and consent. And yes, they could potentially remove artists from the dance-making process.”

This sounds a lot like other wake-up calls from human advocates who suggest that we should be really vigilant about protecting human interests given the wide-ranging and rapidly evolving capabilities of our AI friends. Even truly “human” activities like dancing are being transformed by AI! Keep this in mind, as we see adoption trends continuing this year and in the years to come.