show index hide index
Revolutionary artificial intelligence, for all its advances, can sometimes be anything but sexy in its pursuit of beauty.
Indeed, this technology, based on complex algorithms and objective data, can sometimes lack the subjectivity and human sensitivity necessary to understand the true notion of beauty.
Since beauty criteria are often subjective and culturally variable, AI can struggle to understand them in a holistic and inclusive way.
Thus, although revolutionary and efficient in many aspects, this AI can sometimes miss the very essence of beauty, as perceived by human beings.
The Aesthetic Limits of Artificial Intelligence
I'artificial intelligence (AI) has made major advances in recent years, particularly in image generation. However, the results often show a narrow and stereotypical vision of beauty. Various image generators like DALL-E, Midjourney and Stable Diffusion have been tested to produce portraits of women according to specific criteria. The results reveal a disturbing and homogeneous representation of beauty that raises ethical and aesthetic questions.
The Homogeneity of Bodies
Image generators rely on stereotypical models to create visual representations. When these AIs are asked to generate an image of a « beautiful » woman, the results are surprisingly uniform: thin, young, and mostly light-skinned. This homogeneity of bodies is not only reductive but also reinforces oppressive beauty standards, particularly for women.
Lack of Diversity
One of the most concerning aspects is the striking lack of diversity in the images generated. For example, searches for “normal” women produce 98% fair-skinned women. Racial diversity and signs of age are almost non-existent. You have to ask specifically for images of « ugly women » to see some variety in terms of age and body shape. This failure to reflect a more diverse range of physiques perpetuates ideals of beauty limited and exclusive.
To read LinkedIn : le grand ménage débute, place aux posts authentiques sans IA
The Impact of Biased Algorithms
The reason behind these consistent results lies in the biased algorithms used by AI. These algorithms rely on data from the Internet, where images of thin and young women are overrepresented. Therefore, the AI learns to associate beauty with these specific characteristics, ignoring other forms of beauty.
Social and Psychological Consequences
The lack of diversity and the perpetuation of unrealistic standards have serious consequences. These representations can influence perceptions of beauty and create unrealistic expectations, which can harm self-esteem, particularly among young women. AI, instead of offering an inclusive and varied vision of beauty, on the contrary reinforces oppressive standards.
Potential Solutions
To make AI more inclusive, several steps can be taken:
- Integrate diverse datasets who represent a variety of bodies, ages and ethnicities.
- Set up algorithm checks and calibrations to reduce inherent bias.
- Encourage collaboration with ethics and diversity experts to create more inclusive standards.
These measures would enable the development of AI tools that reflect the richness and diversity of human beauty, instead of reducing it to limited and harmful stereotypes.
