The confluence of A.I. and art takes off
Creativity has long been regarded as a “human-only” capacity, but the recent confluence of artificial intelligence and art is already evolving the future of creative expression as we know it. The following examples illustrate how A.I. art is going from a curiosity to a more mainstream—and even well-funded—creative pursuit.
“PORTRAIT OF EDMOND DE BELAMY”
Last fall, famed auction house Christie’s made history by auctioning art created by A.I. for the first time ever. French art collective Obvious used a GAN, or generative adversarial network, to produce a series of portraits in the style of historic European portraiture. The portraits were created by feeding two concurrent neural networks with portraits spanning the 14th to the 20th centuries until the Generator network could produce an image that fooled the Discriminator network into thinking a human produced it. One of the works, titled “Portrait of Edmond de Belamy,” sold for $432,500, nearly 45 times its high estimate, and officially ushered A.I.-produced art into the global art realm.
“FACELESS PORTRAITS TRANSCENDING TIME”
HG Contemporary, a gallery in the Chelsea neighborhood of NYC, hosted an exhibit titled "Faceless Portraits Transcending Time." All of the prints showcased at the exhibit were computer-generated, created by a machine-learning algorithm that was trained by a computer scientist named Dr. Ahmed Elgammal. The exhibition promoted Elgammal’s fine art econometrics startup (which produced the prints) and showcased the creative potential of A.I. With this exhibit, HG Contemporary claims to have “the first solo gallery exhibit devoted to an A.I. artist.”
AIPORTRAITS.COM FROM IBM WATSON
Researchers at MIT-IBM Watson AI Lab developed an A.I.-powered widget that turns users’ selfies into portraits that mimic classical portraiture throughout art history. Users can head over to aiportraits.com (which is currently experiencing a loading issue due to heavy server traffic at the time of publication) to upload a selfie that is then fed through a GAN, or generative adversarial network. The widget was trained on 45,000 artworks until the GAN’s neural networks could produce a portrait that looks like a user's selfie but is rendered in the style of classical artists such as Van Gogh or Rembrandt.