Researchers at Kyoto University (Japan) have found that people cannot distinguish haiku written by a human from AI-generated haiku. The Mainichi reports.
Participants evaluated the poetry on a seven-point scale across criteria such as “the sense of beauty of the work.”
AI-generated haiku scored 4.56 points, while human-written haiku scored 4.15.
“The results suggest that collaboration between humans and artificial intelligence fosters better creative processes in the creation of haiku,” the study says.
Participants were also asked which haiku were created by the algorithm and which by humans. It turned out that people could not tell them apart.
According to the scientists, AI poetry received higher marks due to the “algorithm aversion” effect. People tend to believe that higher-quality works are created by humans, the researchers added.
Earlier in September, a neural-network-generated painting won a visual-arts competition.
In July, researchers found that users cannot distinguish images created by humans from those produced by neural networks.
Subscribe to ForkLog’s Telegram updates: ForkLog AI — all the news from the world of AI!