University of Oregon researchers say artificial intelligence can identify fake art

Look at the pictures above. Can you tell the real (and valuable) Pollock painting from the worthless fake? An artificial intelligence tool developed by University of Oregon researchers can distinguish between Jackson Pollock and his imitators with 98.9 percent accuracy. (Answer: The image on the left depicts a portion of “Blue Poles,” one of Pollock’s most famous works, while the one on the right shows an imitation.)

(JUNE 28, 2024) Jackson Pollock was a famous artist. He made abstract paintings. Some were worth over $100 million. So many people try to copy his style. This leads to debates when similar paintings are found. Are they real? Or are they fake?

A team from the University of Oregon created an artificial intelligence (AI) tool to help identify real paintings by Pollock. They say this tool is 98.9% accurate. 

“It’s almost like the American dream to find a Jackson Pollock painting in your attic,” says Richard Taylor. He is a professor at the University of Oregon professor. He is also a fan of Pollock's art. “After the excitement, how do you know if it’s a real painting or a good fake? That’s where the machine comes in.”

The tool learns to see patterns called fractals in Pollock’s paintings. Fractals are patterns that repeat in smaller and smaller sizes, like in trees or clouds.

Pollock used a special method to paint. He poured paint on the canvas. This created fractal patterns. It also made his work a good fit for computer analysis, says Taylor.

Julian Smith and Caleb Holt studied at the University of Oregon. They trained a computer to find these patterns. They used nearly 600 pictures. Some pictures were Pollock's real art. Some pictures were imitations. Some pictures were other kinds of art.

Surprisingly, the tool works well but does not show how it makes its decisions. “I can see that it’s providing a result, but I don’t know how it makes all the little decisions,” he says. But that might be a good thing. Taylor says the tool can make decisions without human help. That might help stop fake paintings.

"Even the best Pollock experts can make mistakes,” Taylor says. “If AI can help solve these cases quickly, it will benefit the art world. It’s a great chance for art and science to work together."