A new study investigating how humans differentiate between art created by AI and those created by humans has determined that the majority of individuals primarily can’t tell the difference.
Published within the journal Empirical Studies within the Arts by scientist Harsha Gangadharbatla, the study was galvanized by the sale of “Edmond Delaware Belamy”, an AI-generated portrait by the inventive French studio Obvious. Hailed as “the future”, the painting fetched around 10 times the common value for an average artist at auction, going for $432,500 at Christie’s in 2018.
The ballyhoo around “Edmond Delaware Belamy” wasn’t an isolated prevalence, either. A 2017 study that asked individuals to compare a variety of AI artworks against human art, individuals largely most popular the artworks created by machines.
“For me, the fascinating factor was the role of humans in creating art” Gangadharbatla tells Artnet. “I perpetually assumed there was a soul that the human pours into the work. Once a machine creates the work, how do individuals interpret it?
The study consisted of a survey asking respondents to tell apart human artworks from AI art. The human artworks were created by artists Tom Bailey and Steve Johnson – impressionistic landscapes and geometric abstractions – whereas the computer-created artworks were the work of an AI algorithm.
Of the many people that participated in the survey, most were unable to properly identify more than 1 in 5 landscape works created by AI. Over 75% guessed wrong on the remaining four. Individuals were slightly better at telling apart abstract artworks, signalling a potential association of abstraction with AI and objective art with a person’s creator.
While Gangadharbatla says that AI making artworks is comparatively harmless, he adds that advertising might be a distinct story. “If computers begin coming up with persuasive messages that appear before individuals,” he asks, “What would be the result of that be?”
In more recent news from the AI art world, AI-DA declared a new exhibition of self portraits, set to travel to a show in London. Earlier this year, OpenAI additionally debuted DALL·E, a neural network which will produce freaky pictures supported by written descriptions.
In case you’re still making an attempt to work out whether or not the pictures at the top of this article were created by humans or AI, they are, in order: AI, human, human, and AI.