AI's 'Default Human' Is 97% White Dude CEO: Devs, Time to Code Mandatory Diversity Prompts Before the Bots Unionize?

A recent study has highlighted the issue of bias in AI-generated images, with tools such as DALL-E 2 and Stable Diffusion consistently producing images of white men when prompted with general terms like "CEO" or "software developer". This bias is not limited to occupation, with the AI also perpetuating stereotypes in terms of age, ability, and socioeconomic status. Researchers have found that 97% of images generated by DALL-E 2 are of white men, while only 2% of images generated by the tool depict black individuals. The European Union's ethics guidelines and the World Health Organisation's guidance on AI ethics emphasize the need for transparency, accountability, and inclusivity in AI development. Experts argue that mandatory identity specification could help mitigate these biases, but this approach also carries risks, such as the potential for misuse of collected data. As of 2024, companies like Bloomberg and UNESCO are working to address these issues, with some researchers suggesting that AI-generated human representation may need constraints to prevent harm.