14/06/2026
The problem I have with people saying "AI can't be creative" and "AI will never be creative" and "Don't use AI for art" is not with their intuition. It's with the oversimplification of their message, which I think misses a deeper truth that would make their intuitive point much sharper.
Here are some great papers to read to get an understanding of how academics and researchers are really thinking about this problem.
The academic debate (while not infallible) has largely moved past "is AI creative? yes/no" toward a more nuanced consensus about the problem:
1. creativity is multidimensional and can differ in degree. (different *kinds* of creativity / more or less creative)
2. AI is strong on some axes (combinatorial creativity) and weak on others (transformational creativity, problem-solving creativity)
3. So the REAL problem is diversity collapsing at the population level, rather than lack of creativity at the individual level.
The most fascinating analogy I found in my research was this problem being confronted by copyright law after the invention of the camera. In Burrow-Giles Lithographic Co. v. Sarony (1884), Napoleon Sarony's portrait of Oscar Wilde was reproduced without permission. The defense? A photograph is a mechanical reproduction by a device, not the "writing" of an "author," so it can't be copyrighted. The court's response? It did NOT claim the machine was creative. It conceded the camera was a blind recorder and relocated authorship to the human choices around the mechanical act. (A hostile 1859 Salon review slammed photography as the refuge of failed painters; the camera merely records, no imagination! "It just captures what's already there." This maps almost perfectly onto "it just recombines training data.")
