Researchers debate gender bias and language in AI

Researchers Paloma Piot and Lucía Álvarez recently participated in a debate at the University of A Coruña regarding gender bias in artificial intelligence and the role of the Galician language in computing.

During the session, Piot noted that no current AI model is entirely free of bias. She explained that these systems do not make errors randomly; instead, they systematically affect specific demographic groups or contexts. This occurs because large language models are trained on internet content that contains existing inequalities, hate speech, and biases, which the models then replicate.

The researchers highlighted that bias stems from both the training data and the human decisions made during the development and training processes of these models.