An art gallery uses an AI system that is supposed to check works for their originality in order to detect forgeries or works heavily inspired by others.
The AI analyzes image features, style, coloring, and patterns that it has learned from a large database of known artworks.
After some time, artists and curators complain that the AI frequently rejects innovative or experimental works as not original because it recognizes similarities to already existing styles.
The team wonders: Why does the AI rate many creative and new artworks as unoriginal, even though they are explicitly supposed to be novel?
Question: Why can an AI that evaluates originality based on known patterns mistakenly classify creative innovations as imitations, and what challenges does this pose for the evaluation of art?
Solution follows tomorrow.
Solution
The AI evaluates originality based on patterns it knows from the training data.
Creative innovations often combine familiar elements in new ways or subtly change stylistic features, which the AI interprets as similarity to already existing works.
Since the AI cannot make a genuine aesthetic or cultural assessment, it does not recognize the innovative context, but only statistical matches.
This leads to novel works being mistakenly classified as imitations and thus disadvantaged.
The challenge is that originality is subjective and depends on cultural, historical, and individual factors that an AI can hardly capture.
Result: An AI for evaluating artistic originality reaches its limits because it often does not recognize creative innovations as such, but overlooks or rejects innovations due to its focus on known patterns. For a fair assessment of art, AI systems should therefore be supplemented by human expertise and develop more flexible, context-related evaluation approaches that allow room for creative diversity.