A portrait designed entirely by an AI has sold for a whopping $432,500 (£337,784). This is not an Onion article.
The artwork was sold at New York auction house Christie’s, and was only expected to sell for $7,000 – $10,000. In the end, it actually went for a figure almost 45 times as high as anticipated.
The painting, entitled Portrait of Edmond Belamy, depicts a large gentleman staring mournfully towards the painter. It’s rather blurry and incomplete, but the biggest giveaway that it wasn’t painted by a human is in the bottom right corner where its algorithm is written in place of a signature.
The AI painter is the work of Paris-based AI art collective Obvious, who wrote a two-part algorithm for its AI to create artworks such as this portrait. It’s one of a series of portraits created by Obvious’ AI, all modelled on the fictional Belamy family.
READ NEXT: Architects are teaching AI to print entire cities
The first part of the algorithm is the Generator, which uses a data-set of 15,000 pieces of art from between the 14th and 20th centuries to create an image. Then the second part of the algorithm, the Discriminator, discerns differences between the generated portrait and the dataset’s images, eliminating works that are too far from the dataset until the Discriminator is fooled into thinking a generated image is man-made.
The Discriminator is not yet as good as the human eye at perceiving the flaws of AI artwork, so art experts can still easily spot the distinction between the Portrait of Edmond Belamy and the works it’s based on. To the layman, however, it could easily be a medieval portrait.
AI art market is slowly growing, with one AI capable of drawing realistic-seeming street view images, and another detecting fake artwork in the real world. You also have projects like Vincent, which uses AI to turn sketches into procedural artworks. However, an AI painting an entire original piece that’s sold for so much is an unprecedented move that could beckon the way for more computer-generated canvases.
Image © Obvious
Original publication by Tom Bedford: Source article