Study: Artificial intelligence language models show implicit racism towards African American English

According to new research, large AI language models and OpenAI's version of ChatGPT have hidden racial biases against African-American English.

Here's What We Know

The identified biases are manifested not only in this aspect, but also in the appointment of jobs and the assessment of suspects' guilt. Among law masters, who usually do not show overt racism, certain implicit biases were noticed that became apparent thanks to artificial intelligence.

The authors of the study emphasise the potential risks of using such AI models in areas where they are already being applied, such as business and law. They call for further research and resolution of these issues to prevent further discriminatory manifestations in the future.

Source: Gizmodo