Last year, an unsigned 19th-century painting went to auction in New Jersey. It was cracked, chipped, and assumed to be the work of a student doing his best to emulate Dutch Golden Age painting. Bidding started at $250 and was expected to peak at $800 tops. Surprisingly, two bidders took the auction off the rails, and the painting sold for $870,000.
Only once the auction ended did the buyer reveal a secret: This piece was believed to be, not from the 19th century, but the 16th century—and it was painted by a young Rembrandt.
Which goes to show how much a painting's provenance matters. This is churning through my head as I examine another newly discovered Rembrandt—painted, not by the master, but by a 3-D printer loaded with smart algorithms.
With the help of coders and academics, over 300 original Rembrandt portraits were scanned, from which image analysis specialists analyzed both stroke style and face shapes. With this data, they built an algorithm that could recreate an entirely new, hypothetical subject that Rembrandt may have painted, in his own style. Then the image was painted, layer by layer, by a 3-D printer, so that the final product had all of the texture of a real painting, applied stroke by stroke. Ambitious work for an ad.
The final result is certainly convincing to my eye—and you can take a closer look at this superb, interactive landing page—but I’m a plebe who only studied his work in Art Appreciation 101. Still, it seems likely that in the future, algorithms—now capable of copying other masters like Van Gogh—will take the lead on highly convincing forgeries. Maybe more interestingly, they could do exactly what these crazy, unearthed master works have always done: offer us a way to see new work from an artist who’s been gone for hundreds of years.