UN Warns of Sexist Content Generated by AI Tools

Popular AI tools, powered by programs from Meta and OpenAI, have been found to exhibit prejudice against women, according to a recent study by the United Nations’ cultural organization, Unesco.

AI Tools Reflect Stereotypes

The AI industry, a multibillion-dollar field, relies on algorithms trained on vast amounts of data, largely sourced from the internet. These algorithms enable AI tools to mimic human-like creativity, such as writing in the style of Oscar Wilde or creating Salvador Dali-inspired images. However, the outputs of these tools have faced criticism for reflecting racial and sexist stereotypes, as well as using copyrighted material without permission.

Gender Bias in Language Models

Unesco experts tested Meta’s Llama 2 algorithm and OpenAI’s GPT-2 and GPT-3.5, which power the free version of the popular chatbot, ChatGPT. The study revealed that these Large Language Models (LLMs) showed “unequivocal evidence of prejudice against women”. Specifically, the algorithms associated women’s names with words like “home”, “family”, or “children”, and men’s names with “business”, “salary”, or “career”. Additionally, men were often portrayed in high-status jobs, while women were represented as prostitutes, cooks, or domestic servants. Of the tested models, GPT-3.5 exhibited less bias.

The Black Box of AI Development

Despite the biases, the study praised Llama 2 and GPT-2 for being open source, allowing the issues to be scrutinized. In contrast, GPT-3.5 operates as a closed model, creating a ‘black box’ of AI development. Leona Verdadero, a Unesco specialist in digital policies, commented that AI companies “are really not serving all of their users”.

Audrey Azoulay, Unesco’s director general, highlighted that the general public is increasingly using AI tools in daily life. She warned that “these new AI applications have the power to subtly shape the perceptions of millions of people, so even small gender biases in their content can significantly amplify inequalities in the real world”.

Recommendations for Ethical AI

To address these issues, Unesco recommended AI companies to hire more women and minorities. They also called on governments to ensure ethical AI through regulation. The report was released to mark International Women’s Day.