3 ms·General-purpose large language models outperform specialized clinical AI5 points by doener 4mo ago