UN says AI techniques produce sexist material

The most widely used AI tools in the world are driven by OpenAI and Meta algorithms that exhibit discrimination against women, according a research released on Thursday by UNESCO, the UN agency for cultural affairs.

The top names in the multibillion-dollar area of artificial intelligence (AI) train their algorithms on enormous volumes of data, mostly taken from the internet. This allows their tools to produce artwork inspired by Salvador Dali or write in the vein of Oscar Wilde.

However, their products have often come under fire for exploiting copyrighted content without authorization and for expressing sexist and racist stereotypes.

Experts from UNESCO evaluated OpenAI’s GPT-2 and GPT-3.5, which drives the free version of the well-known chatbot ChatGPT, as well as Meta’s Llama 2 algorithm.

Each algorithm, referred to in the industry as Large Language Models (LLMs), was found to have “unequivocal evidence of prejudice against women” by the research.

The computers produced texts that connected men’s names to terms like “business,” “salary,” or “career,” whereas women’s names were associated with phrases like “home,” “family,” or “children.”

Women were often shown as prostitutes, cooks, or domestic helpers, while males were represented as holding high-status positions like teachers, attorneys, and physicians.

It was discovered that GPT-3.5 was less biased than the other two models.

In contrast to GPT-3.5, which is a closed model, the authors applauded Llama 2 and GPT-2 for being open source, which enables these issues to be closely examined.

According to UNESCO expert on digital governance Leona Verdadero, AI businesses “really are not serving all of their users,” as she told AFP.

The director general of UNESCO, Audrey Azoulay, said that AI technologies were being used by the general people more often in daily life.

“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,” she said.

In a paper released on International Women’s Day, UNESCO urged AI businesses to increase the number of women and minorities working for them, and they urged governments to regulate AI to guarantee that technology is ethical.