Creativity isn’t just for humans. AI-generated works of art have captured the general public’s and even a couple of collectors’ imaginations. However, many have been asking themselves to what extent is AI art the future of modern art?
You may have read on the news that Christie’s became the first company to put AI-generated art on auction. Created through the use of a generative adversarial network (GAN) trained to form new images trained on a dataset of 15,000 portraits painted between the fourteenth and twentieth centuries, the piece sold for $432,500 – over forty times its $10,000 estimate – prompting widespread speculation regarding whether or not AI might be behind the next successful art movement.
Although there’s a 30-year history of artists using algorithms and computation in their practices, this spike in modern AI art, pioneered by artist/software engineers like Robbie Barrat and Google’s Mike Tyka, has occurred within the last 5 years roughly, made doable by recent breakthroughs in machine learning like GANs. The way GANs are trained is to some extent similar to a human brain with created AI paintings increasingly more and more difficult to tell apart from those created by human artists. Also important to the present development of machine learning and art is Google’s DeepDream which was created in 2014. It uses a convolutional neural network to search out and enhance patterns in pictures.
But while technically marvellous, can we conclusively say these artworks are any good? Well, as always, the answer is subjective and depends on the point of view. At this year’s AI-focused Christie’s Art + tech Summit, Jason Bailey, father of art analytics web site Artnome, argued that if what separates our generation from its predecessors is the invention of computers, then AI or machine art is essentially our generation’s most significant art invention. In theory, his statements are reasonable and sensible, however there is still a lot of input required from humans in terms of art selection. For instance, with our current algorithms our art experts diligently select only about 1-5% of the paintings into our AI art collections.
Mike Pepi, who has commented extensively on the intersections between art, culture and technology, remains critical of AI art. ‘I guess I’m simply terribly annoyed with these technical school individuals indiscriminately attempting to use GAN networks in art’ he says. ‘I think there are some artists do that and the results are simply not superb.’
Instead, Pepi points to artists like Agnieszka Kurant and Ian Cheng, whom he sees as using AI as a tool with success – they speak regarding the cultural impact of this technology while not fetishising the results. Cheng’s name is oftentimes associated with an artist who creative person has achieved widespread success through meaningful interaction of his projects with algorithmic processes. BOB (Bag of Beliefs) (2018-2019), as one example. The work comprises, in Cheng’s words, ‘an AI creature whose temperament, body and life script evolve across exhibitions.’ The work’s open-ended narrative, that Cheng has likened to PC games like ‘The Sims’ and reality TV show ‘The Real Housewives’, interestingly has an interactive part – throughout exhibitions viewers are able to observe BOB on a 12-foot-high screen and also can alter its reality by creating interacting with the artwork via an app.
Alongside Cheng, another creative person Trevor Paglen has received mainstream attention for his use of AI technology in art. In September of 2019 he created ImageNet Roulette (2019), an AI tool that labeled user-uploaded selfies in quite a freaky way (e.g. newsreader, smoker, femme fatale) went viral once the racial and gender bias of the datasets on which the algorithm had been trained were discovered. ImageNet Roulette wasn’t Paglen’s first work that looked to reveal how machines ‘see’— his 2017 film Sight Machine has a similar idea—but it absolutely was without a doubt the most successful. Furthermore, apart from providing insight into the typically unobserved process of AI image classification, ImageNet ultimately got rid of 600,000 pictures from its online data bank following accusations that came to light through the use of the tool by Paglen and Crawford’s.
One reason for the success of this piece is that it points to the present gap between the application of AI technology and its limits. ‘It is in some way a mistake to treating AI with exceptionality. We want to be realistic regarding its limits – what it can and what it simply cannot do’ says Lachlan Kermode, Forensic Architecture software package development lead.
Although the above-mentioned works are only a few examples, they suggest that some artists may be well positions to make use of AI technology, however to do that well they may need to critically analyse the cultural implications of a tool’s use. As algorithms progressively have an effect on how exactly we see the world, over time it will be key to build strong AI ethics. ‘It’s crucial that artists are able to get their hands on new technology says Kelani Nichole, a design strategist and the founder of TRANSFER – an L.A.-based experimental media gallery.
Our team at AI Art Shop agree, which is why we are always at the forefront of new AI art technologies, developing custom algorithms that enable us to create stunning exclusive AI created paintings for you. Check out our brand new NFT AI art collection ‘The Source of Inspiration’ here.
