IT House reported on November 18 that according to The Rundown AI, a study conducted by the University of Pittsburgh showed that people cannot distinguish between AI-generated poetry and the works of famous poets such as Shakespeare and Emily Dickinson. Even more surprising, more than 78% of participants gave the AI-generated poems a higher average rating than the works of human poets.
According to IT House, the study surveyed approximately 1,600 participants, who were shown different poems (AI-generated and real works by poets) and asked to indicate which poem they liked. The researchers found that participants preferred the AI-generated poems because they were "more beautiful" and easier to understand. AI-generated poetry received higher ratings in several areas, including rhyme, beauty, and emotional impact , with participants stating that AI-generated poetry had better rhythm and more grace than real poetry.
Even more worryingly, participants mistook the AI-generated poems for human creation 58.5% of the time , and made the correct judgment only 51.7% of the time.
Study author Dr. Brian Porter said: "Our results show that people fail to recognize AI-generated poetry and that they prefer AI-generated poetry to poetry written by humans. In fact, AI-generated poetry is more likely to be judged as written by a human. ’ instead of poetry written by real humans.”
Interestingly, five of the poems rated least likely to be human works were actually the work of famous poets, while four of the poems rated most likely to be human works were generated by AI.