Painter David Salle sends AI to “art school”, teaching machines to see like artists

Artist David Salle has spent two years teaching AI to think like a painter, bridging the gap between traditional art and cutting-edge technology. His experiment represents a significant fusion of artistic sensibility with machine learning, challenging conventional notions of creativity while establishing a collaborative relationship between human artistry and artificial intelligence. This pioneering work demonstrates how AI can become a valuable tool for artists when guided by deep artistic knowledge rather than merely serving as a replacement for human creativity.

The big picture: Traditional painter David Salle ventured into AI art creation after finding most AI-generated imagery lacked the essential qualities that define meaningful visual art.

• Salle identified two critical elements missing from AI art: specificity (precise artistic choices) and intentionality (deliberate creative decisions).
• He recognized that AI’s approach to imagery as an “averaging machine” failed to incorporate the fundamentals of representational painting, particularly the nuanced treatment of edges where shapes meet.

The art education approach: Salle collaborated with technologist Grant Davis to essentially send AI to “art school” through carefully curated training data.

• The training began with a focused selection of works by masters like Giorgio de Chirico, Edward Hopper, and Arthur Dove, each chosen to teach specific artistic principles like perspective, volume, and color use.
• Salle added his own paintings from the 1980s and ’90s to introduce the AI to collage-like, juxtapositional composition techniques.

Why this matters: Salle’s experiment represents a thoughtful integration of traditional artistic knowledge with emerging technology, creating a pathway for AI to serve as a genuine creative tool rather than a mere novelty.

• By identifying and addressing specific artistic shortcomings in AI-generated imagery, Salle demonstrates how human expertise remains essential in guiding technological applications in creative fields.
• The project challenges both technophobia among traditional artists and uncritical techno-enthusiasm by establishing a collaborative relationship between human artistic judgment and machine learning capabilities.

Between the lines: Salle’s approach acknowledges the fundamental differences between digital and physical art while seeking ways to bridge that divide.

• He notes that digital imagery lacks the surface texture and physical presence that gives paintings their object quality in the real world.
• Rather than rejecting AI outright, Salle embraced the technology while maintaining his identity as “a painter, not an engineer” who values the human authorship of handmade objects.