According to a new report, popular artificial intelligence tools are increasingly displaying covert racist tendencies as they progress.
Recently, researchers found that models like OpenAI’s ChatGPT and Google’s Gemini are incorporating African American Vernacular English (AAVE) in their language, leading to potential racial bias in their responses.
Valentin Hoffman, from the Allen Institute for Artificial Intelligence, highlighted how these technologies are being used for tasks like job candidate screening, potentially perpetuating racial discrimination.
In a study, the AI model showed bias by rating AAVE speakers as less intelligent and employable compared to those using “Standard American English.” This bias could have real-life repercussions, such as affecting job prospects or legal outcomes for AAVE speakers.
While ethical guardrails have been developed to address overt racism in AI models, researchers like Hoffman warn that covert racism is still a growing concern as these models evolve.
As the private sector increasingly adopts language models, concerns are mounting about the potential harm these technologies could cause if not properly regulated. Experts are calling for responsible innovation and stricter regulations to prevent discriminatory outcomes in AI systems.
Source: www.theguardian.com
