AI art is causing a stir in the art world and artists, gallerists and collectors are all racing to get to grips with this new technology before some fear it changes the creative industries irrevocably. But what is AI art, why are some artists concerned and others overjoyed and what are its potential uses and repercussions?
What is AI Art?
AI software, such as OpenAI’s DALL-E2, DeepAI, Midjourney, Stable Diffusion and Alexander Mordvintsev’s DeepDream, accepts text prompts from users and turns those commands into a digitally-generated image. The technology uses datasets of hundreds of millions of pairs of images and captions taken from the internet, to “train” the generator. The results get more accurate with each use, as it learns which generated image will most accurately fit the text input. The technology can also be used to “outpaint” existing images, filling in new details or expanding it beyond its original crop (as seen in this ad by Nestlé). Millions of images are being created each day using these websites.
As well as being used in advertising and proving to be a popular software for those not involved in the arts but looking for some online novelty, AI is already making a big impact in the more formal areas of the fine art world. For example, this February an AI-generated image (submitted by Sydney’s Absolutely AI, under the guise of an original photograph) won a major Australian photography competition. AI art has already been selling in major auction houses for high sums and in late 2018 an AI-generated portrait sold at Christie’s for $432,000.
Why Artists are Split
There is a palpable split within the creative community as to whether artists should protest or embrace AI software. The voice of protest is perhaps the loudest online with social media becoming a hub for anti-AI posts. Artists - and particularly digital artists for whom there is a less clear line between their work and a digital forgery - are finding that typing in their own name into the AI command results in being fed near-copies of their designs and styles. To the untrained eye, these AI images could be easily mistaken for original work by the artist. A recent article by the Guardian found that even experts found it hard to differentiate between photographs of original painted artworks and AI-generated versions. These images can be downloaded for free and artists fear that work that is essentially theirs will earn money for others. These artists did not consent for their work to be part of the mammoth datasets on which the software has been trained, so without their permission, their work has been used. Getty Images recently launched the first lawsuit surrounding AI art, claiming that over 15,000 of the 12 million images used in Stable Diffusion’s dataset had been scraped from Getty’s site and some AI-generated images even included versions of the Getty Images watermark. Advertisements for AI copywriting services have writers similarly worried that this technology will lose them jobs and income and is getting people speaking about furthering inequality.
What is AI’s Potential?
AI technology is here to stay and has huge potential for a positive impact on the world. This month, a previously unknown 16th-century panel painting has been attributed to Raphael and studio, using AI technology to analyse the painting’s brushwork to reinforce the attribution to the renaissance master. Art Recognition confirmed that the faces in the ‘Flaget Madonna’ painting were confirmed as over 96% attributed to Raphael. In an art world where up to 50% of art in circulation is estimated to be forged or misattributed, AI technology can be extremely helpful for artwork authentication. Other benefits of AI technology for creatives include its potential for artists to find inspiration for compositions and several artists including Jake Elwes, Mario Klingemann and Soungwen Chung are already incorporating AI technology into their practices.