The human road is long and the sun is setting, but AI has already made astonishing statements? Artificial intelligence (AI) in poetry creation is not a new thing. In this field, early AI mostly relied on simple pattern matching and rule setting, and the generated poetry lacked connotation and emotion. But with the advancement of deep learning and natural language processing technologies, the new generation of AI has been able to capture subtle differences and complex structures in language by learning large amounts of poetry data, thereby generating more natural, fluent, and creative poetry lines. For example, scientists use deep neural networks and reinforcement learning techniques to train AI models to understand and imitate the meter and tone of classical poetry, and even generate poetry lines for specific themes or emotions. Even more astonishing, a study officially published on November 14th stated that readers are no longer able to reliably distinguish between AI generated poetry and human created poetry, and people may prefer AI generated poetry. This interesting phenomenon challenges our traditional understanding of AI creativity and also triggers new thinking about the essence of artistic creation. Scientists from the University of Pittsburgh in the United States have designed a series of sophisticated experiments to explore this issue. Firstly, they invited 1634 volunteers to participate in the test, and each participant received a carefully selected set of 10 poems. This includes five classic works by literary masters, such as William Shakespeare, Byron, Emily Dickinson, and T S. The masterpieces of Eliot and others; The other 5 songs were created by imitating the style of the aforementioned master using the advanced AI model ChatGPT 3.5. The results were surprising: many participants mistakenly believed that AI works were created by humans, and instead held a skeptical attitude towards true human creations. In order to further understand the differences in people's attitudes towards these two different sources of poetry, scientists organized a second experiment. This time, they gathered 696 new participants and asked them to evaluate poetry from multiple dimensions, such as quality, beauty, emotional expression, rhythm, and innovation. The results show that when people know that poetry comes from AI, even if those poems are actually created by humans, their scores on most evaluation metrics will be lower. On the contrary, participants who do not know the source of the poetry are more inclined to give higher scores to AI generated poetry. Scientists believe that the reason for this phenomenon is that AI generated poetry is often more straightforward, easy to understand and feel, which precisely meets the needs of modern readers seeking a simple and clear reading experience. At the same time, readers mistakenly interpret the complexity of human poetry creation as a lack of coherence in AI writing, and underestimate the similarity between generative AI and humans. This discovery not only demonstrates the progress AI has made in imitating human artistic creation, but also prompts us to re-examine the biases that people may have when judging art works. (New Society)