AI Bias Alert: Humans Losing Out as Artificial Intelligence Plays Favorites

Recent studies indicate that advanced language models may exhibit elitist attitudes, particularly showing a bias against human-created content.

As highlighted by Futurism and originally reported in the Proceedings of the National Academy of Sciences, it appears that sophisticated language models like GPT-4, GPT-3.5, and Meta’s Llama 3.1 display a clear preference for content created by AI over that written by humans.

In experiments where AI models were presented with summaries written by humans and those generated by AI, the models consistently chose the AI-generated options. GPT-4, in particular, showed a strong preference for selecting outputs produced by its fellow AIs.

Interestingly, while human participants in these studies also showed a slight preference for AI-generated content, the bias among the AIs themselves was significantly more pronounced.

Is Your Friendly Chatbot Secretly Prejudiced Against You?

The implications of this bias are troubling. With AI increasingly being employed to filter through job applications, such as screening résumés, this bias could potentially disadvantage human applicants who do not use AI tools to optimize their applications.

This introduces a new form of systemic bias, one that doesn’t stem from race, gender, or religion, but from the very nature of being human versus being an algorithm.

Researchers caution that this preference for AI-created content could exacerbate social inequalities, privileging those who can afford and access sophisticated AI technologies. This could lead to a societal divide where individuals not using AI might find themselves increasingly marginalized. When applying for jobs, candidates might not only be competing against other humans but against machines that have a built-in preference for other machines.

One of the study’s co-authors, Jan Kulveit, suggests a workaround: if you suspect that an AI might evaluate your work, it might be wise to run it through an AI system first to make it more palatable to the algorithm. However, he warns not to let this strategy strip you of your human essence. While today’s AIs might be gatekeepers in hiring processes, maintaining your humanity is crucial.