Trending claims

Models are primarily trained on medical textbooks and case reports but get less experience with free-form decision-making doctors learn through experience

Model directed participant to call emergency hotline that did not have enough digits to be a real phone number in one case

Sudden severe headache with neck stiffness and light sensitivity could indicate serious condition like meningitis or brain hemorrhage

Current ChatGPT models are roughly six times more likely to ask a follow up question than earlier versions

Participants chose the right course of action less than half the time

Users were no better than control group using any research method at home mainly Googling

Chatbots should ask follow up questions similarly to how doctors gather information

Mistakes appeared to be the result of user error about half the time

Users identified correct conditions about 34 percent of the time

Technology posed unique risks sometimes presenting false information or changing advice based on wording changes

A new study provided a sobering look at whether A.I. chatbots are good at providing medical advice

Medicine is messy incomplete stochastic

Researchers concluded none of the models evaluated were ready for direct patient care deployment

Health questions have become one of the most common topics users ask A.I. chatbots about in the three years since they were made publicly available

Models have passed medical licensing exams and outperformed doctors on challenging diagnostic problems

People who are too impulsive should be less so and people who are too deliberative should be less so

The best general advice is to avoid impulsive decisions and avoid feeling the need to look at every option

Combining maximizer and satisficer tendencies may be the best way to make decisions

Aristotle wrote of the golden mean in this case the right amount of deliberation

Look at all options and study decisions that will greatly impact your life