Medical AI may look less biased on paper but not in practice, new study finds

Large language models (LLMs) are only as good as the data they learn from. If their training data contains social biases, the ...