Google’s artificial ielligence chatbot in the field of medicine Med-PaLM It is called, in American medical exam He got a passing grade. However, the researchers in their study stated that the responses of this chatbot were insufficie in some cases compared to doctors.
t
Google first iroduced its AI tool to answer medical questions in December 2022 called Med-PaLM. Unlike ChatGPT, this model is not publicly released. According to Google, Med-PaLM is the first major language model to pass the USMLE.
Google artificial ielligence score in the USMLE exam
The passing score of this test is approx 60 perce Is. In their study, Google researchers claimed that Med-PaLM scores on USMLE-style multiple-choice questions 67.6 perce has earned They explain that they developed a new evaluation criterion to ideify and reduce misinformation in their model.
According to a report in May, Med-PaLM 2 Score on the USMLE exam 86.5Perce has obtained that its performance is nearly 20% higher than the previous version of the Google model.
After the launch of ChatGPT last year, a fierce competition between the tech gias in the field of artificial ielligence has emerged. While it has been said many times about the future possibilities and dangers of this technology, we are witnessing new models in various fields, including health.
According to Ahony Cohen, an artificial ielligence expert at the University of Leeds in the UK, one of the biggest problems with these large language models due to their statistical nature is that they provide false information. Therefore, as Cohen suggests, these models should always be considered as an assista.
Earlier this week, it was said that Google’s Med-PaLM 2 artificial ielligence model was being tested at the Mayo Clinic research ceer since April. Although Google has not announced how this model will be used in the research ceer, it has emphasized that it will not pose any risk to paties.
