Artificial intelligence (AI) has yet to crack the code on what makes jokes funny, according to a recent study conducted by researchers at Cornell University. The study involved pitting AI models against humans in various tasks related to humor, such as matching jokes to cartoons and identifying winning captions. The results showed that humans outperformed AI models across the board, highlighting the need for improvement in AI’s understanding of humor.
In a multiple-choice test that required matching cartoons to captions, AI models achieved an accuracy rate of only 62%, compared to an impressive 94% by humans. This significant gap clearly illustrates that AI still has a long way to go in grasping the complexities of humor. Despite this, the researchers believe that AI could be utilized as a tool for humorists in brainstorming ideas, even without a complete comprehension of humor.
To conduct the study, the researchers utilized over 700 entries from the New Yorker magazine’s Cartoon Caption Contest, spanning 14 years. This vast collection served as a comprehensive testbed for assessing AI’s humor capabilities. The findings revealed a noteworthy disparity between AI and human-level understanding of why a cartoon is funny.
The authors of the study propose that AI could potentially be a valuable collaborative tool for humorists, aiding in the generation of ideas. Considering that AI models lack the full understanding of humor that humans possess, humorists could leverage AI’s abilities to assist in the creative process while providing the final touch of humorous expertise themselves.
The study received funding from several sources, including the Defense Advanced Research Projects Agency, AI2, and a Google Focused Research Award. These contributions highlight the importance and interest in furthering the advancement of AI in various fields, including the complex realm of humor.
As AI continues to evolve, researchers and developers will undoubtedly strive to enhance its understanding of humor. While AI’s current abilities may fall short of human expertise in comedic matters, the potential for collaboration between AI and human humorists may lead to exciting breakthroughs in the field.
“Infuriatingly humble tv expert. Friendly student. Travel fanatic. Bacon fan. Unable to type with boxing gloves on.”