An experiment using a custom AI model to evaluate artistic value independent of market context found that the model assigned a seven-figure price to an unknown street artist's work while valuing a Picasso at under $1,000. The project, run by an art journalist with a data scientist and an AI expert, trained a Fine Art Large Vision Model on millions of images and price data. Without metadata like artist name or gallery affiliation, the model's predictions were technically interesting but commercially useless; only when those market signals were added did predictions align with real auction outcomes.
This matters because it exposes the art market's deep reliance on name recognition and institutional validation rather than visual quality. The AI's inability to predict prices from images alone reveals that market value is driven by social and economic biases, not objective artistic merit. The findings challenge the notion that blockbuster artists are simply "better" and highlight how galleries and networks determine success, raising urgent questions about transparency, fairness, and the business realities that art schools often fail to teach.