AI Bias Pt 3

AI bias is a pertinent problem facing AI technologies and human society, but it's also a well-known fact that it reflects the human biases in society which have made their way to algorithms, leading to algorithmic biases. Are these intentional? No, I don't think so. Most developers working on AI systems are not maliciously trying to exclude underrepresented or marginalized groups from the decision-making processes of AI systems. Most of this comes from unconscious biases. And lived experiences. So let's look at the disabled community as an example here and how they're represented in AI images and natural language processing, also known as NLP, which LLMs come from. Now, according to the United Nations, there are about 16% of the world's global population, which represents roughly 1.3 billion people who currently experience significant disabilities. And we also know that not all disabilities are visible. There are many disabilities that aren't physical, from neurodivergence to stammering, depression, schizophrenia, autism, dyslexia, and so on and so forth. And while it's difficult to depict these disabilities in AI systems such as image generators, it is possible to make a concerted effort in increasing the representation of disabled people in AI outputs and ensuring that they're not misrepresented. For example, if you're searching for a disability on an AI image generator, the majority of the images produced are people in wheelchairs. But come on, that's not the only disability out there. When working with AI images or texts, you need to keep a close eye out on the outputs and ensure that underrepresented groups are fairly and equally represented, including people from the disabled communities.