Can machines do comedy? While AI programs can now write essays, generate code, and even mimic art, making people laugh is a whole different challenge because creating humor isn’t just about creativity; it also relies on timing, cultural references, and emotional insight. 

However, a recent study by researchers from Sweden and Germany shows that AI might be catching up in this department, too. The study shows that AI can now create internet memes that are, on average, just as funny as those made by humans. 

“Memes, as a cultural phenomenon, have emerged as a universal language of the internet and are used to express emotions, convey messages, or appropriate and recontextualization familiar elements. This also made them relevant for research as a means of evaluating creative humor,” the study authors note.

The study authors claim this is the first research that examines how human-AI collaboration influences a meme’s appeal and potential to go viral.

“While prior studies have explored the autonomous generation of memes by LLMs, there exists no prior work in examining how human-AI collaboration with a multimodal LLM affects the creativity, humor, and shareability of memes,” they added.

Testing AI’s humor

To test how well artificial intelligence can produce humor, a team of scientists from KTH Royal Institute of Technology (Sweden), LMU Munich, and TU Darmstadt (Germany) carried out what they say is the first large-scale experiment of its kind. 

Their study involved a total of 150 participants divided into three groups (50 each) to explore different creative setups. 

“We conducted a user study with three groups of 50 participants each: a human-only group creating memes without AI assistance, a human-AI collaboration group interacting with a state-of-the-art LLM model, and an AI-only group where the LLM autonomously generated memes,” the study authors explained.

All participants used familiar meme templates, such as the Doge image with Comic Sans captions, the squinting Futurama Fry, and the famous “One does not simply…” line from The Lord of the Rings. The goal was to test how funny, creative, or shareable the memes turned out to be across each group.

Once the memes were made, a different set of about 100 individuals rated them on three fronts: originality, humor, and likelihood of sharing them online. 

Here’s what the researchers discovered: the memes created by the AI alone scored higher on average across the three categories. This means that if you picked a random meme from the AI group, it was more likely to be funny or clever than a random one made by a person. 

However, when the team looked at the best-performing memes—the ones with the highest individual scores—it was the human-made ones that stood out. Meanwhile, memes made by people working with the AI often ranked highly in terms of creativity and shareability, showing that teamwork between people and machines has its own strengths.

“AI scored better, on average, than both human-only and human-AI collaborative memes, but human-created content was strongest in humor. These findings also show that human-AI collaboration in creative tasks is complex. So while AI can increase productivity and produce content that appeals to a wide audience, human creativity is still key for creating content that connects more deeply in certain areas,” the study authors note.

AI may have a tough time creating great humor

The findings offer important insight into how AI and humans can work together in creative fields. While AI shows clear strength in producing content that is appealing and consistent in quality, it struggles with humor that relies on subtle emotions, cultural timing, or lived experience—things that come naturally to people but are hard for algorithms to replicate.

Interestingly, the researchers also found that many people in the human-AI group didn’t make full use of the AI tool. Less than half interacted with it more than once, and only a few used it in a back-and-forth, collaborative way. 

This limited engagement suggests two things: first, perhaps people find it difficult to collaborate with AI when it comes to creating comedy, and second, there’s a need to develop AI interfaces that could encourage a deeper engagement between humans and machines for creative tasks.

Hopefully, future research will focus on solving this engagement issue, helping people stay involved in the creative process while also benefiting from AI’s speed and idea-generation abilities.

The study is published in the journal ACM (Association for Computing Machinery).