A new study has found that readers are unable to distinguish between poetry from artificial intelligence (AI) systems and humans, and even tend to favor poems created by AI.
According to a study published this Thursday in Scientific Reports, most readers can’t even distinguish between classic works and imitations created by artificial intelligence. What’s more: when asked which they prefer, they often choose AI poetry.
Researchers Brian Porter and Edward Macheriff from the University of Pittsburgh in the United States tested the ability of 1,634 participants to distinguish between poetry created by artificial intelligence and poetry written by a poet.
The results of the study
They were randomly presented with ten poems: five written by ten famous poets, including William Shakespeare, Lord Byron, Emily Dickinson, and T.S. Eliot, and five created by ChatGPT3 in the style of these poets.
The researchers found that participants were more likely to say that the poems created by artificial intelligence were written by humans, while the five poems they thought were least likely to be written by humans were actually written by poets.
“More than 78% of our participants gave higher ratings, on average, to poems generated by artificial intelligence than to poems by famous poets written by humans,” said Brian Porter, quoted by the publication.
In the second experiment, 696 participants evaluated 14 characteristics of poems, namely: quality, beauty, emotionality, rhythm, and originality.
Participants were randomly assigned to three groups, where they were informed that the poems were written by humans, created by artificial intelligence, or had no information about the poem’s origin.
According to the study, participants who were told that the poems were created by artificial intelligence gave lower scores on 13 characteristics compared to participants who were told that the poems were written by humans, regardless of whether the poems were actually created by artificial intelligence or written by poets.
However, participants who were uninformed about the authorship of the poems rated the AI-generated poems better than those written by humans.
