Recently, OpenAI's research team discovered that the usernames chosen by users can somewhat influence the responses of ChatGPT during interactions. Although this effect is relatively minor and primarily seen in older models, the findings are still noteworthy. Users typically provide their names to ChatGPT for tasks, making the cultural, gender, and racial backgrounds inherent in those names an important factor in studying bias. In this research, the researchers explored ChatGPT.

Experiments have found that OpenAI GPT-3.5 exhibits racial bias in resume screening. Test results indicate that GPT-3.5 is more inclined towards certain ethnic names, which could impact hiring decisions. The experiments also revealed gender and racial preferences for GPT in different job positions. OpenAI responded that companies usually take measures to mitigate bias when using its technology.

Researchers evaluated the racial and gender bias of GPT-4 in clinical decision-making and found significant biases. The study calls for bias assessments on large language models. The findings were published in The Lancet Digital Health.

AI-generated white faces are now so realistic that they appear 'more real than real', yet they exhibit bias in the representation of other ethnicities. The research found that biases in the algorithm's training data lead AI to favor white features when depicting other ethnic groups, potentially exacerbating online racial bias. The researchers urge public education to prevent misinformation and online scams, highlighting the potential issues brought by AI's 'hyper-realism'.