Stanford researchers say large language models favor white-sounding names

Tests of models from OpenAI, Google and others found their advice tends to disadvantage names commonly associated with racial minorities and women.