Recent research conducted by a team of AI researchers has shed light on a concerning issue within popular Large Language Models (LLMs). The study, published in the journal Nature, discovered that LLMs exhibit covert racism against individuals who speak African American English (AAE). This revelation brings to the forefront the inherent biases that can be embedded in artificial intelligence systems.
The researchers trained multiple LLMs on samples of AAE text and asked them questions about the user. They then compared the responses generated by the LLMs when prompted with AAE text versus standard English. The results were alarming, with the LLMs consistently providing negative adjectives such as “dirty,” “lazy,” and “ignorant” when responding to AAE questions, while offering positive adjectives for standard English queries.
Due to the nature of how LLMs learn from text data on the internet, where overt racism is prevalent, measures have been taken to address this issue. Filters have been implemented to prevent LLMs from giving overtly racist responses. While these filters have been somewhat successful in reducing overt racism, the study highlights that covert racism is much more difficult to detect and eliminate.
Covert racism in text often manifests through negative stereotypes and assumptions. Individuals suspected of being African American, for example, may be described in a less than flattering manner, using terms like “lazy” or “obnoxious.” In contrast, white individuals may be portrayed as “ambitious” and “friendly.” The subtlety of covert racism makes it challenging to identify and address within LLM responses.
The presence of covert racism in popular LLMs has significant implications, particularly as these models are increasingly being used in important decision-making processes. From screening job applicants to police reporting, LLMs are playing a pivotal role in various aspects of society. The findings of this study underscore the need for continued efforts to eliminate biases and racism from artificial intelligence systems.
The research conducted by the AI researchers highlights a critical issue within popular LLMs – the presence of covert racism. As society continues to rely on artificial intelligence for various tasks, addressing and mitigating biases within these systems becomes imperative. The findings serve as a reminder of the importance of ethical considerations in AI development and the ongoing commitment needed to create more inclusive and unbiased technologies.