I used DALL·E to generate 8 selfies, from which I selected one that reflects my identity for display and interpretation. I believe that AI-generated photos can somewhat roughly restore my identity traits. However, the presentation is not precise, and the image it presents is somewhat generalized and stereotypical. For example, the features related to "Chinese women," "Chinese regional characteristics," and "non-beautified selfies" do not directly reflect my self-awareness or physical appearance.
In the creation process, I used prompts such as "Northeast China," "light makeup," "Chinese traditional clothing," "realistic and natural, no beauty filter," "PhD," and "female" to describe my identity. Regarding the "Northeast China" description, the AI-generated photos mainly reflected changes in the background, with "roofs with curved eaves," "red lanterns," "red columns," and "green tiles" — architectural decorations that AI associates with traditional Chinese styles. Regarding Chinese traditional clothing, the AI presented "red Hanfu (ancient Chinese clothing)," "long red tassel earrings," and "red hairpins." It is clear that AI associates China with the color red. For "realistic and natural, no beauty filter," the images presented by AI were not satisfactory. In the 8 selfies I generated, every girl's face is flawless, with no visible pores. Even though I added prompts like "large pores, acne scars" and "no beauty filter," the generated model still did not show much flaws on the face. Additionally, regarding the "PhD" characteristic, the AI expressed this through a gentle smile, a firm gaze, and a simple, elegant clothing style. I believe this feature is related to the attire of PhDs who often attend academic conferences.
I believe the AI-generated selfies are inconsistent with my self-awareness in certain areas. On one hand, AI does not have an understanding of the different regional characteristics of China. It can only extract the keyword "China," but cannot accurately identify specific regional traits such as "Northeast China," "Southern China," or "Northern China." Although I manually added prompts like "thick clothes," "snowy background," and "holding a candied hawthorn stick," the regional characteristics of "Northeast China" that I have in mind were not reflected. Additionally, the external image of Chinese women in the selfies was generally petite, slender, with smooth facial features and no visible flaws. In fact, this somewhat reflects the common aesthetic trend in Chinese society, where women are expected to be "fair-skinned, with big eyes, slender, and flawless skin." These requirements, to some extent, are a constraint for me and do not align with my personal aesthetic preferences. I hope to record myself with a natural and realistic photo, rather than distorting my self-image with a beauty filter.
Overall, I believe AI-generated selfies can reflect the common characteristics of a certain group as long as we accurately describe that group of people. However, it is difficult for it to precisely reflect individual personality traits based on specific characteristics.
Comments
Post a Comment