36 Professionals: The Gender Bias in Generative AI Models
An Exploration of Midjourney’s Image Generation
As society increasingly relies on Artificial Intelligence (AI), it’s crucial to consider how generative AI perpetuates stereotypes and bias. Recently, I was inspired by a campaign by Óticas Diniz (a Brazilian optical shop), which explored how Midjourney generates images of people representing different professions through machine learning algorithms, revealing clear biases.
The video inspired me to explore Midjourney and its generative AI bias. I found biases toward gender and racial stereotypes by using a prompt to make realistic pictures of people in different jobs but changing only the name of the job. When I typed in "Portrait of a startup founder," for example, all of the images that came up were of men. This kind of bias reinforces the prejudices and stereotypes that already exist in society, making it seem like entrepreneurship is mostly for men.
Furthermore, generative AI can perpetuate racial and age stereotypes, with specific terms like “doctor” or “lawyer” often generating images of a White male in a suit. The danger isn’t just about misrepresenting certain groups; it’s also about missing diverse perspectives from people who don’t fit into those stereotypes. These biases, if unchecked, could lead to increasing disparities in industries that rely on generative AI-generated content for hiring or promotions.
To stop this from happening, we need to train neural networks with a sense of ethics and datasets that include people from diverse cultures and underrepresented groups. As we celebrated International Women's Day, it's important to keep in mind how important it is to make sure that everyone is treated equally in all parts of society, including technology. Only then can we build an equitable future for everyone, regardless of their race or gender.
See the complete list of pictures here:
