Top Tech Companies Explore the Humor Challenge for AI Chatbots

Top Tech Companies Google DeepMind, renowned for its work in tackling global issues through artificial intelligence, recently took on a new challenge humor. Researchers at DeepMind, including some with a background in improv comedy, sought to determine if AI could generate genuinely funny content.

In a study published earlier this month, DeepMind’s team asked 20 comedians to evaluate jokes produced by leading chatbots the feedback was not encouraging participants described AI-generated humor as dull, unoriginal, and excessively politically correct one comedian compared it to cruise ship comedy material from the 1950s, but less offensive while some found AI useful for generating a rough first draft, few were impressed with the overall quality of the jokes.

“Comedy, explained DeepMind researcher Juliette Love, co-author of the study, is a deeply human endeavor. It involves personal history, social context, and audience understanding, which pose fundamental challenges for current AI models trained on data from specific points in time.

Comedy, as Juliette Love explains, thrives on human complexity and challenges AI’s historical data limitations, acording to Barron’s Subscription

The AI Humor Race

DeepMind explores AI humor, joined by Elon Musk’s xAI with Grok, a chatbot aiming for humor. Anthropic presents Claude 3.5 Sonnet, emphasizing improved humor understanding. In an OpenAI demo, a voice-enabled GPT model chuckled awkwardly at a dad joke. These developments highlight AI’s ongoing quest for nuanced and genuinely humorous interactions.

“Humor is a tough nut to crack,” stated Daniela Amodei, president and co-founder of Anthropic. “I don’t think Claude is as strong as a professional comedian, but we’ve made definite improvements.”

The Serious Business of AI Humor

For tech companies, infusing AI with humor is a significant goal. They aim to create conversational chatbots that handle complex queries while being engaging enough to foster ongoing interactions in both professional and personal settings.

Amodei explained, “The best colleagues are professional, approachable, and can inject a little humor into a conversation. That’s what we’re aiming for with AI.”

CDK Global cyber incident disrupted auto dealerships.

CDK Global, a prominent software provider for the automotive industry, caused significant operational disruptions for thousands…

Experiments with Chatbot Humor

In personal tests with various chatbots, the humor was underwhelming. When asked for a joke about OpenAI CEO Sam Altman, Grok responded with a dry, “Why did Sam Altman cross the road? To get to the other side of the AI revolution!”

Similarly uninspired were responses from OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude, each offering variations of the classic “crossing the road” joke. Gemini’s attempt was slightly more original: “Sam Altman walks into a bar and orders a round of drinks for everyone. The bartender says, ‘Wow, that’s generous! What’s the occasion?’ Sam replies, ‘Just celebrating achieving AGI… again.’”

The Balancing Act of AI Humor

DeepMind’s Love warned against pushing AI humor too far, as it could increase potential harm to some groups. We must strike a careful balance, she noted. Humor can be polarizing; what is funny to some can be offensive to others. Minimizing this risk might come at the expense of humor.

Human Quote of the Week:

“This Top Tech Companies is special in that its first product will be the safe superintelligence, and it will not do anything else up until then.”
—Ilya Sutskever, Former Chief Scientist at OpenAI

Sutskever, who recently departed from OpenAI, announced his new venture, Safe Superintelligence. He declined to name financial backers or disclose funding details.

Dive into a complete financial experience with The WSJ and Barron’s at 50% off. Get 52 weeks of WSJ Print + digital access, and Barron’s Print + online access. Stay informed and empowered—subscribe to The WSJ and Barron’s today!