Traditional auction house takes on AI artwork

kevin laminto on Unsplash

British auction house Christie's is set to make history with the sale of a portrait created using machine learning algorithms to make it look like human-made.

Technology is enabling new ways to create and experience art. In the Czech Republic, a museum art campaign uses haptic virtual reality gloves to allow people with visual impairments to experience sculptures. Another example is a robot illustrator from Italy that produces personalised artwork on any surface. Now, French art collective Obvious is using artificial intelligence to create paintings that look like they were made by humans.

British auction house Christie’s have prided themselves on being cultural stewards in the art world throughout their 252 year lifespan. Despite this weighty history of tradition, they will be auctioning Obvious’ artwork generated by algorithms for the first time in their history. Richard Lloyd, a Christie’s specialist, said: “AI is just one of several technologies that will have an impact on the art market of the future – although it is far too early to predict what those changes might be.”

The project uses machine learning algorithms – known as Generative Adversarial Networks (GANs) – that can generate images. The algorithms work by analysing a training dataset of images and imitating its features. For example, to create an 18^th century portrait, the creators trained the Generative Adversarial Networks on classical portraits. First the algorithm generates a new image based on the set and then it tries to fool a discriminator into believing that the image is a real-life portrait. When the discriminator can no longer spot any differences, the algorithm has produced a successful image. This new process of creating art is forming a new movement called GANism. Obvious say they will use the proceeds from the auction to develop their algorithm further.