Artificial intelligence has been making headlines for its capabilities in various fields, but can it make us laugh? A recent study was published in the journal Plus one Suggests that AI may have a hand in humor as well. The research showed that AI-generated humor was rated as funny or funny compared to human-generated jokes, even when pitted against The Onion's professional satirical comedians.
Creating humor is notoriously difficult. To be considered funny, jokes need to strike a balance between being surprising and benign. Most people develop their sense of humor through exposure and practice, picking up patterns that make the joke work. The researchers wanted to see if large language models (LLMs), a type of artificial intelligence designed to understand, generate and manipulate human language, could mimic this human skill.
LLMs are designed to use vast amounts of textual data and complex algorithms to build models that are able to predict and generate text. These models learn by processing and analyzing vast datasets, which enables them to recognize patterns, understand context and generate coherent responses to text on cue.
The purpose of this study was to explore whether LLMs can create humor that resonates with people. This question is particularly relevant given the entertainment industry's ongoing debate about the use of AI in creative fields. The study's lead researcher, Drew Gorens of the University of Southern California, noted that recent strikes by Hollywood writers and actors highlight concerns that AI could threaten jobs and creativity in the entertainment industry.
Researchers conducted two major studies to compare the humor production capabilities of AI and humans. They used OpenAI's ChatGPT 3.5 for AI-generated content. The first study focused on comparing ChatGPT's humor with the general public, while the second study compared ChatGPT's humor with professional satirists at The Onion.
In the first study, 105 participants from Amazon Mechanical Turk, an online workforce platform, were asked to complete three humor tasks. These tasks include creating humorous sentences for given acronyms, answering the blanks humorously, and preparing roast jokes in response to hypothetical scenarios. Participants were explicitly told to use their imagination and not copy jokes from other sources.
ChatGPT 3.5 was given the same task, offering 20 responses to each prompt. These AI-generated jokes were then combined with human-generated jokes and evaluated by a separate group of 200 participants, who rated their humor on a seven-point scale.
The AI's jokes were consistently rated high in funnyness in three different tasks: creating humorous acronyms, completing blank statements humorously, and generating roast jokes. Overall, ChatGPT's jokes outperformed the majority of human-generated jokes, with the AI performing particularly well in the roast joke task.
Specifically, ChatGPT outperformed 73% of human participants in the abbreviations task, 63% of human participants in the fill-in-the-blank task, and 87% of human participants in the roast joke task.
In another study, researchers compared AI-generated satirical headlines to onions. They used a convenience sample of 217 students from the University of Southern California. Each student rated the funniness of a mixture of headlines generated by ChatGPT and The Onion, without knowing the source of each headline.
The results showed no significant difference in average humor ratings between the AI-generated headlines and The Onion. Among the top four most rated headlines, two were produced by ChatGPT and two by The Onion. In particular, the top-rated headline was AI-generated: “Native man discovers new emotion, yet can't properly describe it.” This shows that ChatGPT can produce satirical content that is on par with professional writers.
These results show that AI, especially ChatGPT 3.5, has amazing skills in generating humor. Despite lacking emotions and personal experiences, the AI was able to analyze patterns and create jokes that resonated well with people.
“Since ChatGPT itself cannot sense emotions, it can tell novel jokes better than the average human, so these studies provide evidence that you need a good joke to tell yourself.” “You don't have to feel the emotion to appreciate a joke,” Gorns said. .
The researchers also explored whether demographic factors influence humor ratings. It was found that age, gender, and political orientation did not significantly affect participants' preferences for AI-generated versus human-generated jokes. This suggests that the comic appeal of AI was broad and not limited to specific demographic groups.
While the study's findings are interesting, they come with several caveats. For example, comic works were text-based and did not include delivery, which is an important component of humor. AI-generated jokes may not perform well in formats that require timing and presentation, such as stand-up comedy or sketch shows.
The researchers wrote, “This chat GPT can produce written humor at a quality that exceeds the abilities of the general public and equals that of some professional comedians, important for comedy fans and those working in the entertainment industry.” There are implications.” “For professional comedians, our results suggest that LLMs can pose a serious threat to employment. The implications are more positive for those who simply reap the benefits of elevating their daily interactions with a dose of humor. They can approach LLMs for help.
The study, “How Funny is ChatGPT? A Comparison of Human- and AI-Generated Jokes,” was written by Drew Gorns and Norbert Schwarz.