Comparing the performance of ChatGPT GPT‐4, Bard, and Llama‐2 in the Taiwan Psychiatric Licensing Examination and in differential diagnosis with multi‐center psychiatrists | Synapse