ChatGPT and Its Rivals Are Boring and Dull. Making Them Funny Isn't Going to Be Easy

  • Many consider current chatbots to be too dull. The companies behind them are aware of the issue.

  • However, the challenge lies in the limited context provided by the data they’ve been trained on.

No comments Twitter Flipboard E-mail

ChatGPT, Copilot, and Claude always seem coherent and accurate in their responses to our queries, even though they may not be. However, they all share a common trait: their formal and serious tone. This seems to hinder the user experience, and AI companies are working to change this in a significant way.

Chatbots lack humor. In a recent study, researchers at DeepMind asked 20 comedians to share their experiences using chatbots to develop their jokes. According to their feedback, the chatbots were dull, unoriginal, and too politically correct. One comedian mentioned that the AI made jokes that sounded like they were from the 1950s, “but a little less racist.”

Humor is a human thing (for now). As the researchers stated in an interview with Bloomberg, comedy is currently a “deeply human” art form that relies on personal stories and social context. This presents a significant challenge for current AI models, as they’re trained on data from “a snapshot in time,” and chatbots have little context about the situations that they're used in.

Grok, the most casual chatbot available. Since its introduction, Grok, xAI’s chatbot available through X, set itself apart from others with its sarcastic manner of responding. X owner Elon Musk specifically positioned it as a more enjoyable alternative to its competitors.

Funny chatbots. Anthropic recently launched Claude 3.5 Sonnet, claiming that it can better understand when users use a humorous tone in their queries. Daniela Amodei, the co-founder and president of Anthropic, stated: “I don't think Claude is as strong as a comedian that you would pay money to see, but I think we have definitely improved.” Meanwhile, at OpenAI, the new version of its model, GPT-4o, also demonstrated the ability to tell jokes and engage in more “human-like” conversations.

The future of chatbots seems less bland. The leaders of these companies believe that being able to use funnier and less formal language will be important for increasing user engagement. As Amodei stressed: “If you think about the people you probably like working with the best, they’re professional, they’re approachable, they’re high integrity, right?… They’re honest, but they can inject a little bit of humor into a conversation with you.” That’s the goal here.

However, chatbots shouldn’t miss the mark so much. It’s clear that user interactions with these chatbots will be more successful if the conversations feel natural. Overly formal and robotic responses, on the other hand, might be off-putting. While it’s crucial for the experience to be enjoyable and even fun, there’s something even more important: Chatbots shouldn’t make so many mistakes.

This is a particularly challenging issue to resolve, especially considering that current chatbots are based on complex probabilistic models and lack a true understanding of their responses. Therefore, it’s essential to always verify their answers, as they may be completely fabricated or contain inaccuracies.

Related | AI Services Are Everywhere, Ushering in the Return of Subscription Fatigue

Home o Index