Humans are bad at distinguishing AI works from human art, but prefer the latter - study

Bowling Green State University

Researchers at Bowling Green State University conducted a study to determine whether people can distinguish artwork created by generative artificial intelligence from human works without prompting.

Here's What We Know

In the experiment, participants were shown images, some of which were drawn by artists and some of which were generated by AI. The task of the subjects was to determine the authorship of each picture. At the same time, they were not informed that some of the works were created by AI.

As the results showed, people correctly identified the source of the artwork only 50-60% of the time, and even then they were unsure of their choice. However, further analysis of the responses revealed that participants were consistently more positive about the works they attributed to human authorship.

According to the researchers, people subconsciously perceive some differences in AI-generated images because of the uncanny valley effect. But they cannot explain what exactly the difference is. Nevertheless, human art evokes more positive emotions and a sense of connection for them.

According to the scientists, the study demonstrates the capabilities of modern generative AI models to create realistic visual art. However, people are still emotionally drawn to works made by humans.

Source: TechXplore