Human-AI Collaboration in Performing Arts

Project Mission

Over the past decades, an increasing number of performing artists and technologists have brought robots to the center stage in their performances. By presenting a series of projects looking into the possibilities of combining AI with performing arts, the mission of the project is to challenge the idea that live performance is a specifically human activity and explore different forms of human-AI collaboration.

Introduction

AI and human collaboration in the theater is the future. This section focuses on the intersection of AI and performing arts, exploring various application scenarios of human-AI interactions in the theater.

“Theatre is about the human expression on stage, and it’s about the communication and empathy between the actors and the audience.” An audience can easily feel and relate to the actor’s emotions through their acting, voice, dance, and movement. Can ‘emotionless’ robots be good performers on stage? AI has taken roles such as comedy improvisers, storytellers, actors, dancers, and choreographers, disrupting the traditional two-way human actor-human audience interaction. I think the dynamics between robot actors and their human counterparts, and the robot actor-human audience interactions are particularly interesting. These experiments also raise questions, including who is the author of the work? How to define acting? Where lies the line between artificial and natural?

The first two works, Hello Hi There and Improbotics discuss the human-AI collaboration to explore the meaning of creativity and spontaneity in improv comedy. Using AI to provoke emotions and build connections with the audience is witnessed in an immersive theater project, Frankenstein AI. The last two works in the dance world are choreographing for machines and by machines.

Augmenting Artistic Improvisation: AI as an improviser

Hello Hi There by Annie Dorsen

Hello Hi There (2010) by theater director Annie Dorsen, cast two chatbots to imitate a human conversation, using the famous 1970s television debate between the philosopher Michel Foucault and linguist/activist Noam Chomsky as the textual database and algorithmically generate an improvised discussion.

HumanMachine Artificial Intelligence Improvisation at Camden Fringe, August 2017

Improbotics is a comedy show fusing improvised theatre with artificial intelligence and telecommunication technology. The machine ‘performer’ is AI-based chatbot A.L.Ex (which stands for Artificial Language Experiment). It was trained on more than 100,000 films.

Figure 1: High level system diagram of A.L.Ex. Source: Mathewson, Kory W. “Humour-in-the-loop: Improvised Theatre with Interactive Machine Learning Systems.” (2019).

The machine can provide inspiring ideas. AI acts like a “completely drunk comedian” who says unusual things (that less likely to be generated by a human) on stage and create the “accidentally funny” moment. The function of AI in improv comedy is basically to challenge the human improviser to make the conversation funny.

AI-based Immersive Theater: AI as a “mirror”

Frankenstein AI: a monster made by many at Sundance, 2018

Frankenstein AI: a monster made by many, developed and produced by Columbia University School of the Arts’ Digital Storytelling Lab. The story is basically AI searching for information about what it means to be human. AI asks the audience questions about humanity, aggregates human emotions, send the answers to sentiment analysis, and reflect back to the audience.

The objective of the project is to “center the needs of humanity within the larger conversation around AI”.

Creating movement: AI as dancers and choreographers

Living Archive: A tool for choreography powered by AI

Living Archive, an AI-powered tool for choreography, was a collaboration between Google Arts and Culture Lab and Studio Wayne McGregor. The team used the archive from Wayne to train the algorithm, and AI invents original movement. According to McGregor, AI can learn and create a particular style of a dancer.

“Create your own dance”. Source: https://artsexperiments.withgoogle.com/living-archive/

The tool is open for the public to explore millions of movements from the archive and create their own choreography.

Discrete Figures, Gray Area Art and Technology Center in San Francisco, 2018

Created by Rhizomatiks, Elevenplay, and Kyle McDonald, Japanese production Discrete figures present a human/machine duet. Human dancers performed alongside with AI dancers (a 3D body projected on stage). The tool that generated movements is a neural network called dance2dance.

“Don’t be a robot on stage!”

People always describe those terrible performances as ‘robotics’, meaning that the performers are ‘emotionless’, and they fail to convey their feelings and engage with the audience.

Machines might not be able to express their feelings and emotions as humans do. But it doesn't mean this would take away the theatrical experience. In the context of human-machine collaborative art, AI has a function of a ‘mirror’ reflecting things about humans based on what human audience, actor, or programmer ‘feed’ them, and human audience complete meaning-making.

It is noteworthy that the collaboration between AI and theater also helps researchers design machines that can understand human behavior better. Acting methods and some practice techniques such as reaction, repetition, and breaking away from mirroring, offer useful insights into multi-agent and human-interaction systems.

--

--