Shruti Gautam
TLDR: As AI Tools Get Smarter, They’re Growing More Covertly Racist, Experts Find
Key Points:
- Popular AI tools like ChatGPT and Gemini hold racist stereotypes about speakers of African American Vernacular English (AAVE).
- AI models are more likely to assign AAVE speakers to lower-paying jobs and recommend harsher penalties for criminal defendants using AAVE in court statements.
In a recent report, experts have found that as artificial intelligence (AI) tools advance, they are becoming more covertly racist. Large language models like ChatGPT and Gemini were found to hold racist stereotypes about speakers of African American Vernacular English (AAVE), an English dialect spoken by Black Americans. The AI models were significantly more likely to describe AAVE speakers as “stupid” and “lazy” and assign them to lower-paying jobs.
Researchers worry that AI models could penalize job candidates for code-switching between AAVE and standard American English, affecting their employability. The study also found that AI models recommended harsher penalties for criminal defendants using AAVE in court statements, raising concerns about bias in the legal system.
While ethical guardrails aim to regulate the content AI models can communicate, the study found that as language models grow, covert racism increases. Experts like Avijit Ghosh emphasize the need for regulation in the use of AI technologies, especially in sensitive areas like employment and criminal justice.
The report highlights the growing market for generative AI, projected to become a $1.3tn industry by 2032, raising concerns about the potential harm AI models could cause if technological advancements outpace federal regulation. The call for regulation echoes concerns raised by leading AI experts like Timnit Gebru regarding biases and discrimination in AI systems.
