AI is destroying art — and maybe saving it?

We’ve all seen the many ways AI is affecting the art world, from a barrage of AI-generated slop to its infiltration of auction houses and the upper echelons of the industry.

But what if AI could also help to keep art’s integrity? It might be possible.

There are currently three conventional ways to authenticate a work of art, per Wired:

• Connoisseurship: The oldest and most common method, it involves experts physically examining an object and providing their opinion.
• Provenance research: follows the documented history of an artwork — through archival materials, letters, gallery listings, etc. —  to discern when the object actually existed.
• Forensic testing: where conservators use tools like carbon dating, X-rays, and infrared spectroscopy to see if a piece of art contains authentic elements.

But now, a fourth way to authenticate art has arrived: AI.

AI art experts

Zurich-based Art Recognition says it has completed 500+ authenticity evaluations by using two types of artificial neural networks.

• The company compiles image datasets for each artist containing reproductions of verified artworks for its AI to train on.
• It also includes contrast sets of known forgeries, works by anonymous artists, and AI-produced pieces to help its algorithms spot fakes.
• After training, the AI can recognize an artist’s brushstrokes, color variations, and compositional elements in previously unseen work.

The bot has been busy: It might have identified a piece by famed artist Peter Paul Rubens (though it’s at odds with scholars) and potentially debunked a fake Van Gogh.

As Wired points out, AI could help art experts — not replace them. It’s a valuable tool, like an X-ray, that authenticators can have in their pocket for tricky cases.

Plus, it could save us all some money. While we might not be waving our auction paddles at Sotheby’s, forgeries are everywhere.