While current AI systems can tell jokes, they often struggle to grasp the relevance of humor in context, leading to misinterpretations or taking jokes literally, particularly in conversations involving sarcasm.

Humor is complex, and its multifaceted nature poses challenges for both humans and AI alike, according to Julia Rayz, a professor and associate Department Head of Computer and Information Technology at Purdue Polytechnic Institute.

“When people communicate, humor plays a very important part,” she said. “So with advancements in conversational AI we are trying to make interaction as natural as possible and humor of course is going to be a part of that.”

Rayz believes that while AI may not currently excel at understanding humor, it holds immense potential. By gradually enhancing AI’s grasp of humor, conversations with AI can become significantly more enjoyable for people. This underscores the ongoing efforts to make AI more akin to human communication. Rayz emphasizes how integrating humor into AI systems is a pivotal step in this process.

“Computers can already tell jokes especially with generative AI playing such a role,” she added. “They can tell jokes, they can read jokes from books, they can come up sometimes with their own jokes. The question is how relevant is it for a human. It has to fit in the conversation, it has to fit in the topic. But also, it would be nice if computers could understand what humor is when humans are talking about and making jokes.”

Developers are interested in adding humor to AI to make chatting with artificial intelligence feel more like talking to a real person, which makes it more fun and comfortable for users. Rayz suggests that even though AI might not fully get jokes yet, teaching it humor can still make our everyday chats with AI better.