New research confirms it: the creativity of artificial intelligence (AI) is a myth. Although current generative AI models may appear to be autonomous creative agents, analysing their imaginative process step by step reveals that their creative abilities are not genuine. This is the conclusion of a new study published in Advanced Science and led by an international team of experts from the Cognition and Brain Plasticity research group at the Institute of Neuroscience (UBneuro) of the University of Barcelona, the Institute for Biomedical Research (IDIBELL), the Computer Vision Centre (CVC-UAB) and the Vienna Cognitive Science Hub.
The study, focused on visual creativity and imagination, began in 2024 during a workshop organized by the Fundació Èpica - La Fura dels Baus, whose work aims to promote interdisciplinary collaborations between science, technology and art.
Following this workshop, the experts devised an innovative methodology to study creativity. They prepared a visual-creative imagination task based on abstract stimuli, and compared the creative performance of an image-generation AI model, with and without human guidance, with that of two groups of people: visual artists and the general population (non-artists). To ensure the drawings were comparable, the AI model was trained using the creative productions of the human participants and was given a more or less elaborate prompt depending on whether it was evaluated with or without human guidance.
Unanimity among the evaluators: human productions are more creative
A group of people and two AI systems were responsible for assessing the degree of creativity of the drawings according to five criteria: liking (i.e. to what extent they liked the drawing), vividness, originality, aesthetics and curiosity. In all cases, the results were clear and unequivocal: the visual artists received the highest score (most creative), followed by the general population, the human-guided AI model, and, by a wide margin disadvantage, the unguided AI model.
"Although the AI model was trained with the creative productions of human participants, it showed a poor performance in the production of creative images; in fact, it did even worse when it was deprived of human assistance," explains expert Xim Cerdá-Company, a researcher at IDIBELL and the CVC-UAB and co-leader of the study.
Studying creativity as a process, not just for results
For the research team, the study's contribution to both AI and cognitive science is manifold. On the one hand, it highlights the need to employ a diverse range of measures and models when investigating a process as complex and multifaceted as creativity. "Currently, AI creativity is valued almost exclusively through verbal creativity tasks, skewing the results and even presenting AI as a creative agent. With a different approach, and by directly assessing the imaginative process from ideation to execution, we have shown that this is not true," continues Cerdá-Company. "Creativity must be studied as a process, not just focused on its results," he adds.
