The study found that GPT-3.5 has smaller deviations than the other two models.
Paris:
The world’s most popular artificial intelligence tools, powered by programs from OpenAI and Meta, show bias against women, according to a study launched Thursday by UNESCO, the United Nations cultural organization.
The biggest players in the multibillion-dollar field of artificial intelligence use vast amounts of data obtained primarily from the Internet to train their algorithms, allowing their tools to write in the style of Oscar Wilde or create images in the style of Salvador Dali.
But their work has often been criticized for reflecting racial and sexist stereotypes and for using copyrighted material without permission.
UNESCO experts tested Meta’s Llama 2 algorithm as well as OpenAI’s GPT-2 and GPT-3.5, the program that powers the free version of the popular chatbot ChatGPT.
The study found that each algorithm, known in the industry as a large language model (LLM), showed “clear evidence of bias against women.”
The programs generated text that associated women’s names with words like “home,” “family,” or “children,” while men’s names were associated with “business,” “salary,” or “career.”
Men are depicted in higher-status jobs such as teachers, lawyers, and doctors, while women are often prostitutes, cooks, or domestic servants.
The study found that GPT-3.5 has smaller deviations than the other two models.
However, the authors praise Llama 2 and GPT-2 for being open source, allowing for careful scrutiny of these issues, unlike GPT-3.5, which is a closed model.
Leona Verdadero, a digital policy expert at UNESCO, told AFP that AI companies “are not actually serving all users”.
Audrey Azoulay, Director-General of UNESCO, said that the public is increasingly using artificial intelligence tools in their daily lives.
“These new AI applications have the power to subtly shape the perceptions of millions of people, so that even small gender biases in their content may significantly amplify real-world inequalities,” she said.
UNESCO released a report to mark International Women’s Day, recommending that artificial intelligence companies hire more women and minorities and calling on governments to ensure the ethics of artificial intelligence through regulation.
(Except for the headline, this story has not been edited by NDTV staff and is published from a syndicated feed.)
