The researchers concluded that the safety filters used by AI to prevent generating offensive content also hindered them from producing material common in comedy writing. The AI's inability to deviate from the norm and surprise the audience, which is crucial in comedy, was also highlighted. Despite some comedians finding AI useful for tackling a blank page or generating a structure for a comedy sketch, the consensus was that the quality of the AI's comedic material was lacking.
Key takeaways:
- Google DeepMind researchers studied the experiences of professional comedians using AI in their work, finding that while AI models were effective at simple tasks like structuring a monologue or producing a rough draft, they struggled to produce original, stimulating, or funny material.
- The comedians reported that they enjoyed using AI models to write jokes, but didn't feel particularly proud of the resulting material, describing the models' jokes as bland, generic, and boring.
- The safety filters that OpenAI and Google use to prevent models from generating violent or racist responses also hinder them from producing the kind of material that's common in comedy writing, such as offensive or sexually suggestive jokes and dark humor.
- The study also exposed the AI models' bias, with several participants finding that a model would not generate comedy monologues from the perspective of an Asian woman, but it was able to do so from the perspective of a white man.