During an experiment, researchers from Google DeepMind discovered that AI-based chatbots are unable to write humorous jokes.
A group of experts enlisted the help of twenty professional comedians by having them write stand-up routines using chatbots such as ChatGPT and Bard. The comedians were then asked to assess the results and provide feedback. The experiment took place online and at the 2023 Edinburgh Festival Fringe.
It has been discovered that LLMs are unable to generate humorous text. “Tasteless,” “very bad,” and “boring” were the adjectives used to describe the end product by stand-up comics. One individual remarked that it was like “cruise comedy stuff from the 1950s, but a little less racist.”
The research found that while AI can write the framework and some of the content for lengthy stand-up routines, it is missing the human touch that gives the material its comedic punch. Individuals “add so much more nuance, emotion and subtlety through their life experiences.”
Robots can’t grasp the meaning of irony, sarcasm, or dark humor.
He decided I wanted to commit suicide, so he didn’t write me any dark things, even though many of my things have dark moments. “So he just stopped giving me anything,” one of the experiment’s participants noted.
Cultural context and audience were stressed by the comedians.
According to one, “my social context would change,” meaning that a comedian’s comedic style in India would differ significantly from their UK counterparts.