A recent study conducted by German researchers has revealed that artificial intelligence (AI) chatbot, ChatGPT 3.5, appears to be limited in its ability to produce new jokes. In their research, the investigators discovered that more than 90% of the jokes generated by the chatbot were mere repetitions of a previously recorded 25 jokes. The study was conducted by Sophie Jentzsch and Kristian Kersting from the German Aerospace Center and the Technical University Darmstadt.
The researchers asked ChatGPT to produce a joke 1,000 times, and they found that all responses were grammatically correct, with nearly all outputs containing just one joke. Prompting the chatbot in various ways did not result in any noticeable differences in its output, except for instances when asked for more than one joke, which resulted in 1,008 responses of jokes in total.
Some Reddit users observed that ChatGPT repeated the same joke about a tomato and salad dressing multiple times when asked to deliver a joke. The researchers also found that this joke was ChatGPT’s second most commonly repeated output, appearing 122 times. The chatbot’s most frequent joke was, “Why did the scarecrow win an award? / Because he was outstanding in his field, which was produced 144 times.
The researchers concluded that ChatGPT relied on mixing elements from previous jokes it had learned instead of creating new jokes from scratch. The chatbot exhibited an understanding of stylistic wordplay and double meanings, but it struggled with sequences that didn’t fit into its learned pattern.
In conclusion, while ChatGPT 3.5 may be impressive in its ability to produce grammatically correct responses, it appears to lack creativity when it comes to crafting new jokes. This study demonstrates the limitations and challenges that still exist in creating AI technology capable of producing unique content that can match human creativity and humor.
