The AI and fine art collaboration is starting to bear real fruit. Case in point: This past week, one of the 11 portraits of the fictional “Bellamy” family, all generated by generative adversarial networks (GANs), sold at a Christie's auction for $432,500, about 43 times more than the $10,000 that was originally estimated. But while the computer model that created the Bellamy portraits was aiming to be as good as the 14th- to 20th-century paintings it was trained on, the Rutgers Art & AI Lab’s “creative adversarial network” (AICAN) generated paintings that were deliberately a departure from its training data, the idea being to come up with something original and authentic. As a result, about 75 percent of the people who viewed the AICAN-created paintings thought they were the work of humans, a study revealed. Whether any of these paintings are actually art or not, however, is still a point of contention, even among AI art practitioners. Nevertheless, art buyers aren't deterred: an AICAN painting sold for $16,000 late last year, which, in light of the Christie’s sale this past week, seems like a bargain.
sign up for our newsletter