Non-expert readers prefer poems written by artificial intelligence over those created by humans. A study conducted by researchers at the University of Pittsburgh examined how a general audience responded to poetry generated by AI.
The study presented participants with poems written by well-known English-language poets like Geoffrey Chaucer and Walt Whitman, alongside AI-generated imitations created using ChatGPT 3.5. To their astonishment, participants were more likely to mistake the AI-generated verses for human-authored works. In fact, the study, published in Nature.com, revealed that AI poems had a roughly 75% chance of being deemed human-written compared to those penned by actual poets.
In a further twist, the research found that AI poems were ranked higher in terms of overall quality than their human counterparts. The authors of the study attribute this to AI's ability to produce clear, accessible poetry that resonates with casual readers. Non-expert readers appreciate the straightforward communication of moods and themes found in AI works.
The complexity and ambiguity often present in human-written poetry, the study claims, may not appeal to those who do not engage in in-depth literary analysis. While traditional poems reward deep reflection, AI-generated poems tend to communicate their messages more directly and unambiguously, making them more accessible to a broader audience.
Despite the findings, poet Joelle Taylor expressed skepticism to The Guardian about AI's role in the world of poetry. Taylor, who won the TS Eliot Prize for her collection C+nto & Othered Poems, emphasized the importance of human experience in poetry. "A poem is more than an algorithm," she stated. "It is meaning, empathy, revelation, inversion, dissidence, passion, and surprise." Taylor believes AI is producing works based on historical norms, which may not represent the full spectrum of what poetry truly is.
The study authors also raised concerns about the implications of AI in creative fields, noting the difficulty many readers had in distinguishing between human and machine-generated poems. They suggest that governments may need to consider new regulations to ensure transparency in AI-generated content.
