At the Mayo Clinic
Google has already begun testing its artificial intelligence in a US hospital. A chatbot based on the Med-PaLM 2 language model has been undergoing testing at Mayo Clinic since April.
Recall that the PaLM 2 language model underlies the Google Bard chatbot. The Med-PaLM 2 version, as the name implies, is designed specifically for medical applications.
The company has been testing a chatbot based on the Med-PaLM 2 language model since April
According to Google itself, such AI could be especially useful in countries with “more limited access to doctors.” At the same time, like other similar language models, it is still far from ideal and suffers from a number of problems typical for such developments.
During the study, physicians found more inaccuracies and outdated information in responses provided by Google Med-PalM 2 than in other physicians’ responses. Yet in almost every other measure, such as showing evidence of reasoning, consensus-supported responses, or no evidence of misunderstanding, Med-PaLM 2 performed at about the same level as clinicians.
Med-PalM 2 is in its early stages and is certainly not meant to replace physicians right now.
