Google DeepMind researcher Piotr Mirowski, during his spare time, trained his solo stand-up comedy skills with a microphone. Sharing the findings, he explored how comedians can leverage AI in their craft.
In this study, he collaborated with 20 professional comedians who were already familiar with incorporating AI in some aspects of their work, such as generating content or drafting initial speech scripts. Using tools like ChatGPT or Gemini, they crafted dialogues that resonated well with the performers and seamlessly integrated them into their routines without revealing that the entire process was AI-driven.
The results were quite diverse. Many comedians acknowledged AI’s ability to swiftly generate ideas, but they often found the material lacked authenticity when translated into live performances. The most successful application was in outlining storytelling frameworks, although tweaking and refining details were still necessary.
Another challenge was the narrative coherence that AI produced. Some described it as reminiscent of old-school comedy plots from the 60s and 70s, lacking originality and at times appearing mundane. This could stem from both ChatGPT and Gemini filtering out results to avoid offensive or inappropriate content. Mirowski suggested that comedy may be a demanding task for AI as it requires pleasing a varied audience, making it potentially despised by everyone.
He emphasized that comedic artistry originates from human experiences rather than data sources, posing a significant challenge for AI. The ambiance during a performance also plays a crucial role, further complicating matters for artificial intelligence.
TLDR: Google DeepMind researcher collaborated with comedians to explore AI’s impact on stand-up comedy, with mixed results highlighting the challenge of infusing creativity and authenticity into AI-generated content.
Leave a Comment