Studies confirm that ChatGPT, the world's most popular AI language model, has a number of biases in its responses. The AI model training involves a preference for copyrighted works in the public domain, with too many sci-fi and fantasy books. ChatGPT, currently the world’s most popular AI-based language model from OpenAI, is not immune to bias in the information it provides, according to a study from researchers at the University of California, Berkeley.
Their dataset on which it is trained and bases its answers, the bias inherent in that data also finds its way into its results. The researchers wrote in their study that OpenAI models have incorporated a large number of different types of copyrighted materials, and this process is influenced by “the frequency with which passages of those books appear on the web