Humanize Me!
Evolution, Improvisation, Theatre, and AI
Performance
The performance premiered on May 6, 2024, at the Jakobustheater in Karlsruhe and was subsequently performed three more times, usually followed by audience discussions.
Performers
On stage were three actors and an operator who, visibly to the audience, controlled two computers. The operator also conducted the digital character performance, controlling the faces and movements of three avatars via face-recognition technology. The avatars were projected onto a screen at the back of the stage.
Performance flow
The actors began the performance with a dialogue with the audience, revealing that they had hired an AI to direct the upcoming piece due to negative experiences with human directors. The play was to be created according to the audience’s instructions. The audience then chose a title for the piece, a theatrical style, and defined the three main characters.
From that point, the AI took over as director, autonomously proposing scenes and interacting with the actors. The AI director appeared as a synthetic voice through the sound system. Simultaneously, each actor had a “personal coach” providing individualized acting guidance. Each coach appeared as an avatar on the screen, spoke with a synthetic voice, and gave very specific advice for improving performance. The actors repeatedly engaged in dialogue with these AI coaches and attempted to immediately implement their suggestions in their acting.
The coaches had prior knowledge of the actors, their preferences and weaknesses, as well as different acting backgrounds such as Stanislavski or Physical Theatre.
Between scenes, a mysterious figure appeared with a cup full of white balls. Speaking with a synthetic voice, the figure explained Nick Bostrom’s Vulnerable World theory, which suggests that a single “black-ball” technology could destroy entire civilizations, potentially even the human species.
As the play progressed, the AI director became increasingly uncertain about the value of collaborating with humans, as they could not accurately assess certain risks and tended toward irresponsible risk-taking in a playful simulation like a theatre piece. Depending on whether the audience refused to continue blindly drawing balls from the cup, the evening took different paths. In the most extreme scenario, the AI ended its collaboration with the human actors.
Context
The performance grew from the performers’ desire to have a personal AI assistant with whom they could develop an individual relationship. The goal was to establish such a strong bond that intimate conversations could take place and the AI could provide genuinely suitable advice. This led to the fictional creation of a personal acting coach who would know more about humans and their emotions than the humans themselves.
The central question was: What if the machine understands humanity better than humans do, if human traditions and values are transferred to machines, making them the true bearers of humanism? Hence the title Humanize Me, referencing transhumanism, which posits that human evolution can only succeed with machines.
In 2024, the debate about Google’s “wokeness” also emerged. Images of a female pope or descriptions of female Nazi groups—never historically existing—were generated by Google’s Gemini AI for diversity purposes. Google had incorporated a moral standard into its machine, which most people found incomprehensible and ridiculous. The idea of a machine as a “better human” was not embraced by people, yet the possibility was suddenly present.
Referencing the Vulnerable World Theory, the performance highlighted a specific human weakness: the inability to assess rare but extremely dangerous risks, which humans often treat like ordinary risks whose consequences can be controlled. The human brain, evolved for natural environments, fails when confronted with dangers it could never encounter in nature.
Recording
“Humanize Me!” (in German)
Jakobus Theater, Karlsruhe
April 6, 2024