Science demands both intelligence and creativity, yet research on automating science rarely considers the latter. We take a creativity perspective and present a case study on Google’s AI Co-scientist, a system that helps researchers generate novel hypotheses in biomedical domains. We analyse the AI Co-scientist using concepts from Computational Creativity, organised around four perspectives: product, process, producer, and press. Our analysis shows that the AI Co-scientist already demonstrates several creative traits, even if not framed that way by its authors. An analysis of the arts through the same four perspectives highlights that not only the creation but also the communication in the arts depends on subjective and emotional understanding, making them harder to automate. We conclude that Computational Creativity can provide useful models, theories, and evaluation methods for advancing automated science.
Toivonen et al. (Thu,) studied this question.