Robot theatre is an experimental project where the theater is made by an artificial intelligence named John, who writes the play himself and creates the production. The project will be completed in collaboration with artists, researchers, entrepreneurs and students of Tallinn University.

The plan is to direct artificial intelligence to do theater – from writing a play to staging a performance and finally performing on stage.

The Czech writer Karel Capek first introduced the word “robot” created by his brother Josef in 1921, and he does so in the theater. The show “RUR” (“Rossum’s Universal Robots”) is about a man-made robot who eventually attacks a human. Two facts seem more than meaningful here:

1. “Robots” were born on the stage, not to wash the floors

2. The “robot” will turn 100 this year and must be celebrated with dignity


First, we feed the neural networks certain models of dramaturgy: first, Chekhov’s plays. Later, depending on the result, we will also administer Beckett’s plays, Kivirähk’s texts or opera librettos instead.

At the beginning, we test different models based on Chekhov’s texts (different neural networks and different text input methods or ways to further train neural networks). We will continue to work with the best “play” and bring it to the stage with only robots, robots and human actors, or with human actors and robotic suffixes. The latter would guide actors not only in words, but also in stage movement, emotions and other activities.

When creating a production, we also use neural networks in different ways: we feed production schemes, stage design principles and visuals into the machines. The future depends a lot on the result of the work of neural networks so far. An important question is based on which logic to train neural networks for this purpose and how to achieve the substantive integrity and poetic effect of the production. The aim is to give the machines maximum autonomy, but it will probably be necessary to intervene at times. For example, we want to maintain a certain simplicity in the visual aspect – so that when allocating roles, robots do not imitate humans or human visions of the robot, but rather take the form of an object or a more abstract phenomenon – such as a chair, cloud, wall or even an invisible deity.

To direct the work towards a poetic and emotionally influential performance, a team of people is ready to direct the work of the robots, looking for suitable ways to provide appropriate input to the neural networks.


We hope to get closer to understanding the possibilities of using artificial intelligence in the creative process, more precisely: in interactive drama. On the stage, we hope to see an unprecedented genre, unexpected technical solutions, wit, but above all an emotional and poetic performance.

Experimenting with different types of neural networks and machine learning algorithms provides important information – what and on what basis to use networks in different phases of creative work; what to keep in mind when entering raw data; what are the possibilities of using physical and autonomous robots in performances? This would be a necessary knowledge – the use of autonomous agents in creative applications, be they interactive films, educational applications, or perhaps productions, is an indispensable area for further research and creative work.