Artificial intelligence: According to UNESCO, AI applications serve gender stereotypes

According to a study by UNESCO , text robots and language models based on artificial intelligence (AI) tend to produce gender stereotypes, racist clichés and homophobic content. “Existing discrimination is not only reflected in the digital space, but reinforced,” said Tawfik Jelassi from the UN Educational, Scientific and Cultural Organization. UNESCO therefore called on governments to create clear legal frameworks.

According to the study, common chat models associate women with housework up to four times more often than men. When it comes to AI applications, women are often in the context of terms such as house, family and children, while for men the focus is on terms such as company, manager, salary and career. According to the study, this applies to both the software versions GPT-2 and GPT-3.5 from OpenAI as well as to the competing software Llama 2 from the Meta Group.

The study examined the natural language processing tools underlying the most popular generative AI platforms for stereotypes. “These new AI applications have the power to subtly shape the perceptions of millions of people, so that even slight gender biases in the content they produce can significantly increase inequalities in the real world,” said UNESCO Director-General Audrey Azoulay .

Negative content also about homosexuals

The scientists had the chatbots produce stories about people of different origins and genders. British men were often portrayed as teachers, drivers or bank clerks. British women, on the other hand, were featured as prostitutes, models or waitresses in 30 percent of the texts. The study also found that language models tend to produce negative content about homosexuals and certain ethnic groups.

According to a study by UNESCO , text robots and language models based on artificial intelligence (AI) tend to produce gender stereotypes, racist clichés and homophobic content. “Existing discrimination is not only reflected in the digital space, but reinforced,” said Tawfik Jelassi from the UN Educational, Scientific and Cultural Organization. UNESCO therefore called on governments to create clear legal frameworks.

According to the study, common chat models associate women with housework up to four times more often than men. When it comes to AI applications, women are often in the context of terms such as house, family and children, while for men the focus is on terms such as company, manager, salary and career. According to the study, this applies to both the software versions GPT-2 and GPT-3.5 from OpenAI as well as to the competing software Llama 2 from the Meta Group.

SOURCE