AI Walks Into a Bar: The Quest for Artificial Humor

In a world where artificial intelligence (AI) can beat chess grandmasters and drive cars, researchers may face their toughest challenge: teaching machines to tell a good joke.

As AI systems grow increasingly sophisticated, tackling everything from medical diagnoses to autonomous vehicles, a new frontier has emerged: humor. This isn’t just about programming machines to recite knock-knock jokes. It’s about creating AI that can understand, generate and deliver contextually appropriate humor in real-time interactions with humans.

Teaching Machines to Laugh

Several strategies for training AI in humor have emerged, Pedro Domingos, professor emeritus of computer science at the University of Washington, told PYMNTS. “Fine-tuning the models on collections of jokes, cartoons, humorous essays, and books, etc., available on the web. Explaining to the models what’s funny and appropriate vs. not, and prompting them in various ways until they produce something to our liking. Training the models to produce funnier and more appropriate humor by having humans rate their output accordingly.”

However, he cautioned, “None of these are a guarantee of success, though, and humor is still one of the harder things for AI models to do successfully.”

The process involves complex data analysis, Sean Vosler, founder of, told PYMNTS. “Researchers utilize large datasets containing examples of humor, such as jokes, comedic scripts, and humorous social media posts, to train these models. These datasets help AI learn patterns in language that we associate with humor, including wordplay, timing, context, and cultural references.”

Yet, the challenge goes beyond mere data processing. “Humor is particularly challenging for AI because everyone finds different things funny,” Dave Edwards, co-founder of Artificiality and former product manager at Apple, told PYMNTS. “For an AI to succeed at being funny, it has to understand what its audience finds funny overall and understand if the current context is appropriate for humor.”

The Business of AI Humor

The potential applications of humor-enabled AI are vast and varied. “I think there’s room for humor in just about any interaction between a chatbot and human, just as with interactions between humans,” Domingos suggested. “For example, it can make customer service more friendly and help defuse frustration.”

However, he also warned of potential pitfalls: “I could easily see attempts at chatbot humor falling flat or even alienating customers.”

In the realm of customer service, humor could transform interactions. “It can be used to create more engaging conversations and less transactional interactions,” Vosler noted. This approach could potentially defuse tense situations and enhance overall customer satisfaction.

The entertainment industry also stands to benefit. “Imagine an AI that not only gives therapeutic advice but also makes you laugh a little,” Ghazenfer Mansoor, CEO and founder of Technology Rivers, told PYMNTS. “There’s something uniquely human about using humor to ease tension and connect with others.”

The context in which an AI chatbot is deployed also plays a critical role, Joscha Koepke, chief product officer at Connectly, an AI-powered conversational commerce company, told PYMNTS.  “For example, an AI customer support agent handling flight cancellations does not benefit from being funny, and people will likely react negatively to humor in stressful situations,” Koepke said. Conversely, an AI chatbot recommending fashion products can enhance user experience with a touch of wit.

Cultural differences and language nuances present another layer of complexity in developing humorous AI. “To address this, developers use localized datasets and cultural sensitivity training for AI models,” Koepke said. 

As the field progresses, researchers and developers continue to push the boundaries of what’s possible. They’re not just aiming to create AI that can tell jokes but AI that can engage in witty banter, understand and use sarcasm, and perhaps even develop its own unique sense of humor.

“In the big picture, adding humor to AI is about removing walls between people and machines,” Mansoor reflected. “This way, our digital talks feel more normal and like talking with real humans.”

The quest for artificial humor continues, promising a future where our interactions with AI might just be a bit funnier. Mansoor said, “Stay tuned; the future looks very funny.” Yet, as the experts remind us, the path to truly funny AI is paved with complex algorithms, cultural sensitivities, and the elusive nature of humor itself.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.