A new paper found that large language models from OpenAI, Meta, and Google, including multiple versions of ChatGPT, can be covertly racist against African Americans when analyzing a critical part of their identity: how they speak.

Read more...