Artificial Intelligence (AI) is a tool that many of us use in our daily lives. One of its many uses includes generative AI. From generating texts to realistic images, and even videos simply from word prompts, it is redefining the way that society is consuming media. While this poses benefits in the creative scene, it poses an important question: How does it reinforce stereotypes?
Generative AI models rely on vast data derived from the internet and data providers. These include social media posts and news articles among many other sources. However, human biases are also one of many things that are captured by these models.
To examine the magnitude of biases in generative AI, Bloomberg generated thousands of images using Stable Diffusion – a text-to-image model that creates hyper-realistic images from word descriptions. The model was tasked to generate images for 14 different jobs of various income levels, and the results were revealing:
- High-paying jobs were dominated by subjects with lighter skin tones while darker skin tones were associated with jobs such as “fast-food worker” and “social worker”.
- Most occupations in the images generated were dominated by men, except for low-paying jobs.
- Men with lighter skin tones dominated high-paying jobs like “judge”, “CEO”, and “lawyer”.
These findings reveal how generative AI can mirror and amplify societal biases
So, what is its impact on society?
As of now, generative AI is being used by advertisers. However, this could grow into other forms of media that shape our daily perceptions. Perpetuating stereotypes through imagery may unintentionally reinforce harmful assumptions and create barriers for minority groups in areas such as career advancement, education or social representation.
An even more troubling possibility is the use of generative AI in fields like criminal justice. Imagine if AI were tasked with creating suspect images—doing so could amplify existing prejudices, leading to further marginalization of already overrepresented groups in the justice system.
As generative AI becomes increasingly woven into the fabric of our society, its potential to shape our perceptions and reinforce biases cannot be overlooked. While this technology holds incredible creative and functional potential, the unintentional perpetuation of stereotypes is a pressing challenge we must address. To build a more inclusive society, recognizing these biases is the first step.
With responsible use and vigilance, generative AI can definitely be used as a tool that uplifts rather than limits.
Reference:
https://www.cigionline.org/articles/generative-ai-tools-are-perpetuating-harmful-gender-stereotypes/
https://www.uva.nl/shared-content/uva/en/news/press-releases/2021/11/which-stereotypes-are-ingrained-in-ai-language-models.html
https://www.washingtonpost.com/technology/interactive/2023/ai-generated-images-bias-racism-sexism-stereotypes/
https://www.bloomberg.com/graphics/2023-generative-ai-bias/ https://longreads.com/2023/10/18/how-ai-reduces-the-world-to-stereotypes/