A new study suggests that artificial intelligence can have "creative thinking," once again surpassing humans.

In a study conducted by doctoral students at the University of Arkansas, 151 individuals were compared to ChatGPT-4 to measure divergent thinking.

Divergent thinking is often considered a sign of creativity and imagination, as it involves a thinking process used to generate ideas and explore possible solutions to concrete or abstract problems.

In this study, it was found that the multimodal language model created by OpenAI provided more "original and nuanced answers" than human participants.

The first task involved people's creative use of everyday objects, such as forks or ropes. The second task, called the consequence task, required individuals (as well as their AI counterparts) to imagine the outcome of fictional scenarios. The third task, called the "divergent association task," required 10 nouns that were "as semantically distant as possible."

The researchers then observed the length of responses and the semantic differences between words. They found that the ChatGPT-4 model "was more creative and nuanced than humans in each of the different thinking tasks, even when controlling for fluency of responses."

While a comprehensive understanding of AI creativity has yet to be widely researched, this study marks the first exploration of its capabilities.

The researchers have some "notes of caution"

In a press release published on the University of Arkansas website last Friday (March 1), the authors said, "It is important to note that the measures used in this study are indicators of creative potential, but engagement in creative activities or achievements is another aspect of measuring one's creativity..."

"AI is different from humans, it does not have agency," "It relies on the assistance of human users. Therefore, unless enhanced, the creative potential of AI will remain stagnant."