Measuring Machine Imagination: Evaluating AI-generated creativity

Abstract: With the advent of open-source generative artificial intelligence programs, the adage “What is art?” is again amidst the turmoil. Access to Text-to-image software online has flooded the internet with “art” created by users, unsettling the creative communities. The debate of whether the art generated by this software can be classified as art is a burning question of today. Although the community is divided, there is no denying that AI is here to stay and will continue to expand into previously untapped areas in the years to come. Considering this, this research tries to compare the assessment of the aesthetic values of art generated by artists and AI software. An experiment was conducted with the students by giving them an excerpt from the book “Invisible Cities,” which describes a fictional world rich in imagery. Simultaneously, these texts were fed into Midjourney to generate various interpretations. After evaluating multiple models of aesthetic assessment, through parameters of usability and intent, a deep neural network was selected. The assessment results were then understood regarding each case and contrasted. Although there were no significant differences in the “aesthetic score” of the artwork, an underlying pattern emerged that these artworks projected, which was directly linked to the complexity of the provided text. Practical Implications: The research aimed to analyze the creative potential between the Human Mind and the AI mind, which in turn is the amalgamation of our minds collated. Keywords: Artificial intelligence; AI art; creative collaboration; Human AI art.