AI Masterpieces: But is it Art?
While there’s no doubt that the quality of the results in style transfer is outstanding, many were left with feelings that the technique left little room for the concept of art itself, even calling it “… more of a parlor trick than the next revolution in fine art.”
Image synthesis using generative adversarial networks
Robbie Barrat, now a researcher at Stanford, had the idea of using generative adversarial networks (GANs) to enable algorithms to synthesize art. And a wickedly powerful idea it is. A GAN is a package of two “adversarial” networks. The first, the generator, tries to make images that the second, the discriminator, will accept. The discriminator’s role is to differentiate between generated images and real images, on which it was trained. When it “fails” an image that’s passed to it by the generator, it provides feedback from which the generator can learn and make another attempt.
Robbie fed a GAN with 10,000 nude portraits and let the two networks try to fool each other. And then something remarkable happened. The generator learned how to fool the discriminator not by accurately creating nude portraits, but by creating strange and surreal images that are just convincing enough to slip under the discriminator’s radar. They are also a simply fascinating mashup of discombobulated blobs and features, fascinating enough to possibly be called art. Robbie’s Twitter feed is alive with such images, as is his Github profile, where he has generously placed much of the enabling technology. Here’s a montage of synthesized nudes:
Paradoxically, an earlier experiment with a set of landscape paintings didn’t go wrong like the nudes did—it generated a bunch of quite convincing landscapes and is arguably far less “creative” as a consequence.
25 October 2018 marked a first for the auctioneers Christie’s and the world, when, in London, the first ever auction occurred of a work of art created by an algorithm.
It’s all the work of a Paris-based collective called Obvious, who created it using the same GAN principles that underly Robbie Barrat’s work. Depicting a fictitious member of an equally fictitious family, Portrait of Edmond Belamy was expected to fetch between 7000 and 10,000 euros but the final price reached stunning $432,000.
Is this a watershed moment for AI art? Or is it a frothy piece of marketing? Speaking in late 2018 Robbie Barret seems pretty certain: “The work isn’t interesting, or original” he is quoted as saying, “They try to make it sound like they invented or wrote the algorithm that produced the works … but people have been working with low resolution GANs like this since 2015.”
Without having a dog in the fight, it seems to us, and with all due respect to Robbie, that if this marks the beginnings of a similar movement from all of the top-tier auctioneers, then AI art might very well have reached the tipping point that propels it into the mainstream. And not a moment too soon. However, AI holds lots of other, more practical, opportunities, which we, at AI consulting and development company Iflexion, continue to explore.
Bio: Yaroslav Kuflinski is an AI/ML Observer at Iflexion. He has profound experience in IT and keeps up to date on the latest AI/ML research. Yaroslav focuses on AI and ML as tools to solve complex business problems and maximize operations.
- Picking an Optimizer for Style Transfer
- Beyond the Fence, and the Advent of the Creative Machines
- Machine Learning for Artists – Video lectures and notes