New study tests medical advice accuracy of ChatGPT and finds AI improving over time, but ethical concerns remain

Study tested accuracy of ChatGPT-3.5 and GPT-4 in answering 284 physician-prompted medical questions. Found high rates of accuracy and completeness overall.

ChatGPT answers improved in accuracy when retested 11 days later, suggesting the AI learns over time.

GPT-4 showed slight improvements in accuracy and completeness over GPT-3.5.

Chatbots have potential to provide efficient medical advice, but raise concerns about privacy, misinformation, and replacing human medical roles.

Advances in AI chatbots are fascinating but must be implemented carefully and ethically in healthcare.