From the Fine Art Connoisseur July/August 2024 Editor’s Note:
AI As Our Ally
I have written before about how wary we should be of artificial intelligence (AI) getting too mixed up with the making of art. I stand by those comments, and I sense that ever more people in our field have their guards up, too.
One aspect of AI that thrills me, however, is its capacity to help us better understand historic works of art. An exciting example came to light last year when it was announced that significant portions of a previously unrecognized painting of the Madonna and Child with Saints Elizabeth and John the Baptist were painted by the Italian Renaissance master Raphael (1483–1520). This discovery rolled out in the ideal manner: some humans got the hunch first, and then the computers proved them right.
To be specific, Dr. Larry Silver, professor emeritus of art history at the University of Pennsylvania, proposed that the 16th-century painting was made by Raphael’s studio assistants, with the all-important faces of the Madonna and Baby Jesus attributable to the master himself. Then thorough examinations of both the pigments and the provenance were undertaken by the London and New York-based firm Art Analysis & Research, which gave it the green light.
Finally, the painting was subjected to scrutiny by the Zurich-based firm Art Recognition, which has patented an AI system for authentication through analysis of brushstrokes. First, it designed a deep convolutional neural network (DCNN) trained on a dataset containing images of all of Raphael’s known paintings, as well as a contrast set of fakes, comparables, and contemporaries. This allowed the system to identify feature series such as brushstrokes, variations in color, and high-level composition elements, which it proceeded to seek in 16 individual sections and across the painting as a whole. Indeed, it found that much of the surface was painted by artists other than Raphael, but the face of the Madonna was a 96.57% match with Raphael and Jesus’s came in at 96.24%.
“Through brushstroke artificial intelligence, we offer objectivity and accessibility to
clients, which has been missing in the field of art evaluation for many years,” says Dr. Carina Popovici, founder and CEO of Art Recognition. “Art history, provenance, chemical analysis, and other methods are all critical to the full understanding of an artwork, but attribution decisions should not be left solely to the subjective human expert’s eye.”
The Illinois cabinetmaker and artist Tony Ayers discovered this painting in an English antiques shop in 1995. The seller had acquired it from another dealer based in Kentucky, who had purchased it from the local Sisters of Charity. Later research showed it had been donated to them in 1837 through Bishop Joseph Flaget, and thus it is now called “The Flaget Madonna.” Before that, it was believed to be part of the Vatican collections. Ayers spent much of his adult life studying the painting with various scholars and scientists; after his death in 2023, his widow and friends continued his pursuit.
Art authentication has always involved patient sleuthing, and it’s thrilling to know that AI can aid in our efforts by diving more deeply — and more dispassionately — into the evidence. As long as humans remain in control of the technology, there is reason for us to expect many more memorable discoveries in the future.
What are your thoughts? Share your letter to the Editor below in the comments.
Download the current issue of Fine Art Connoisseur here.
Sign up to receive Fine Art Today, the free weekly e-newsletter from
Fine Art Connoisseur magazine.
 ...){kind=link}