When exposed to terms common in different racial dialects, large language models make racist assumptions about people from particular racial groups, even without explicitly knowing their race.
Scientists have discovered that common AI models express a covert form of racism based on dialect — manifesting chiefly against speakers of African American English (AAE).
In a new study published Aug. 28 in the journal Nature, scientists found evidence for the first time that common large language models including OpenAI’s GPT3.5 and GPT-4, as well as Meta’s RoBERTa, express hidden racial biases.
Click here to read the rest of this story.