Google's Med-PaLM 2, an Artificial Intelligence (AI) tool intended to answer medical information queries, has been undergoing testing at the Mayo Clinic research hospital and other facilities since April. Med-PaLM 2 is a version of PaLM 2, the language model that supports Google's Bard, announced at Google I/O in May this year.
Google's belief that its updated model could be particularly beneficial in countries with "more limited access to doctors." Med-PaLM 2 was trained on a curated set of medical expert demonstrations. This approach is expected to make it superior in healthcare conversations compared to general chatbots like Bard, Bing, and ChatGPT.
Despite this, a study Google published in May highlighted that Med-PaLM 2 still faces some accuracy issues common to large language models. Physicians reportedly found more inaccuracies and irrelevant information in Med-PaLM 2's answers compared to those provided by other doctors. However, in most other metrics, such as demonstrating reasoning, providing consensus-supported answers, or showing correct comprehension, Med-PaLM 2 performed comparably to actual doctors.
The customers testing Med-PaLM 2 will have control over their data. The data will be encrypted and inaccessible to Google.
According to Greg Corrado, Google's senior research director, Med-PaLM 2 is still in its early stages. Corrado expressed his hesitancy about including the tool in his own family's "healthcare journey," but he believes that Med-PaLM 2 "takes the places in healthcare where AI can be beneficial and expands them by 10-fold."
