Zeroing in on the origins of bias in large language models

When artificial intelligence models pore over hundreds of gigabytes of training data to learn the nuances of language, they also imbibe the biases woven into the texts.