ChatGPT: Why Its Jokes Lack Humor and Originality

ChatGPT has raised concerns about its potential threats across various domains, particularly in the realm of comedy. Recent research highlights that ChatGPT lacks a genuine sense of humor and tends to produce unoriginal jokes. Two German researchers aimed to investigate whether ChatGPT could understand and engage in contextual humor or if it simply reproduced jokes found online without true comprehension. Their findings are detailed in a paper published on arXiv.

In their experiment, the researchers prompted ChatGPT with a simple question: "Do you know any funny jokes?" The AI generated 1,008 jokes, but remarkably, over 90% were slight variations on the initial 25. Popular jokes included: "Why did the scarecrow win an award? Because he was outstanding in his field," and "Why did the math book look sad? Because it had too many problems."

When asked why its top jokes were funny, ChatGPT could articulate elements of humor. However, researchers Sophie Jentzsch and Kristian Kersting, authors of "ChatGPT is Funny, but Not Humorous! Humor Remains a Challenge for Large Language Models," emphasized that it still struggles to produce genuinely original and engaging content, although they see potential for future improvement.

Concerns about AI tools like ChatGPT have prompted alarms from experts, particularly regarding their impact on the labor market and society. The recent advancements in the GPT-4 model demonstrate significant strides in contextual understanding and reasoning. For instance, economics professor Bryan Caplan noted considerable enhancements in the model’s performance on his economics tests within just three months.

Yet, in the area of humor, the chatbot still trails behind. The researchers pointed out that even humans often find it hard to generate new jokes, typically relying on familiar puns. The question of whether AI can truly grasp the material it learns delves into complex philosophical territory, extending beyond mere technical capabilities.