On the last Monday of August, Jason Allen won first prize in a digital art competition in Colorado. Controversy arrived however, when it came to light he had utilised an AI software that rendered the painting for him. Midjourney, a text-to-image AI generator, can create images from command prompts entered into its search bar.
So, rather than using a digital brush or other tool to actually create the artwork, Allen typed in what he wanted his painting to look like, and waited for the results. When he revealed how we had made the art (which he had by then printed onto a canvas), the internet broke into an explosion of debate, as usual.
Some of the louder voices on twitter commented that they were ‘watching the death of artistry’. @OmniMorpho lamented that both creative and high-skilled jobs are in danger of becoming obsolete. People criticised cheaply produced art meant to ‘be consumed in bursts of a few microseconds as it glides by on the infinite feed.’
Midjourney is one of the leaders in text-to-image AI, a software that can create images from prompts given into a search bar. Open AI, the current leader in the field, has recently released its updated software, DALL-E 2.
The software uses billions of parameters to understand the prompts being asked of it. DALL-E 2 is more than just a database of pictures however. It has been shown to blend concepts when creating its images, and demonstrates other key aspects of creativity usually only associated with humans. For instance it can pass the Raven Matrices test, designed to test human’s non verbal reasoning.
Despite its incredible potential, so far Open AI, Midjourney, as well as their open sourced counterparts, have been making a storm on the internet because of their usefulness to create memes and other abstract online jokes. The avocado armchair was a particular hit.
Aside from a few technical issues that need smoothing out, there are more serious concerns. In an attempt to prevent violent or sexist imagery from being produced, DALL-E 2’s data was filtered. This then resulted in a bias towards male over female images (it is thought this is due to women more frequently being sexualised than men, an unpleasant look in the mirror for our society). A scientist, programmer or judge will be depicted as a white man, an assistant or a ‘lazy student’ will appear as a woman – often overtly sexualised. Worse still, typing ‘Homeless person’ will bring depictions mostly of black men. Open AI gives a disclaimer that despite attempting to clean up the data, its AI might ‘reinforce or exacerbate societal biases.’
Furthermore, there are serious ethical concerns with text-to-image generators like DALL-E. Professor Toby Walsh expressed his fear of synthesised photos being used maliciously with the help of this software. In the world of fake news, having an intelligent AI that can render anything from Donald Trump joining the Communist Party to Taylor Swift sieg heiling, will be very dangerous, particularly as the technology gets more advanced.
It will come as no surprise that Allen received a lot of backlash for his work at the Colorado State Fair, with people arguing he did not make the piece himself, and was not honest about how the art was made – until after he won. Allen did not apologise, questioning why the difficulty making a piece should have any impact on its reception. He also rebuked the claim he did not have any part in the creative process – he said about ten percent of the project was himself editing the AI’s work.
Allen argued people were only criticising him because they were scared of being replaced, which is probably true.The first time there was a digital artwork created, painters no doubt complained. Likewise photographs caused the downfall of painting portraits. While these concerns are justified, there is no real way to prevent the tide of progress.
Matt Borrs criticised The Atlantic for using an AI generated piece as the cover, rather than an illustrator, arguing it was merely a tool ‘to make billionaire’s richer’, at the detriment of real artists. Although the concern of traditional artists is no doubt justified, some thought may be had for the designers and developers of the AI technology. Although not as glamorous as a painter or cartoonist, they presumably enjoy getting some recognition themselves. This opens up an ongoing debate on art, and specifically what should be recognised as art?
Can Cy Twomley or Marina Abramović paint like a Dutch master? Can David Guetta play an instrument? In video game creation, thousands of hours go into making vast, impressive, artificial landscapes and even whole worlds, but they don’t usually receive the same recognition as a traditional artist. Is this the end of art, or just its evolution?
Dr Oliver Bown debated AI in art, arguing the clunky, humorous look of the generators (particularly the free to use DALL-E Mini), will create a subculture through social media, and become an art form in its own right. DALL-E, a software whose name is taken from a Surrealist artist and a cartoon character from an entertainment conglomerate, sums up the bizarre mix of cultures.
DALL-E 2 is not perfect – yet. On its website Open AI seems happy to point out mistakes made by the software, arguing it ‘makes it more human’. The ethics around anthropomorphising of technology is a talk for another day; but the issues with this software are largely irrelevant. As it evolves, it will become more intelligent. The vast improvements from DALL-E Mini, to DALL-E, to DALL-E 2 shows us its current lack of intelligence will soon be negligent.
The selling of search engine prompts by users of the software, in order to get quicker and more accurate results, is another step in the direction of digital art. In the world of instant gratification, where we can work from home and have our food delivered to us at the touch of a button, an art platform where you can simply Google your masterpiece into existence has taken off.
The future may see the introduction of rules for AI, different categories in competition for AI and traditional art, for example. The question of compensation or recognition for artists whose work has been studied free of charge by the AI is also a topic of contentment – as of now their work is being studied for the AI to function without their consent. As with all technology that is new and groundbreaking, the rules and regulations are taking their time in catching up. What these developments will mean for traditional artists and traditional art, digital or otherwise, remains to be seen.