Google is testing its artificial intelligence chatbot in hospitals

The Wall Street Journal reported this morning that the Mayo Clinic Research Hospital is testing an artificial intelligence program called Med-PaLM 2. The program is a type of language model, meaning it can understand and use language like humans, so it can be used to analyze medical data or answer medical questions. The Med-PaLM 2 model is a variant of another language model called PaLM 2, which was introduced by Google in its conference called Google I/O in May of this year. This language model powers Google’s chatbot called Bard.

According to Hoshio, an article in the Wall Street Journal says; According to an internal email, Google believes that the Med-PaLM 2 language model could be particularly useful in countries with “less or limited access to doctors.” Med-PaLM 2 is a computer program trained using a specific set of information related to a medical specialty. Google has designed this program to be more suitable for healthcare conversations compared to other more general chatbots such as Bing, ChatGPT, and even Google’s own chatbot; Bard, better work. Google believes that because the program is trained using medical expertise, it will be more advanced, more specialized and better equipped to understand and respond to the specific needs of people discussing healthcare-related topics.

The article also cites research that Google made public in May (pdf) showing that Med-PaLM 2 still suffers from some of the accuracy problems we’ve become accustomed to seeing in large language models. . The physicians who conducted the study found that the responses provided by these linguistic models, Med-PaLM and Med-PalM 2, provided more irrelevant and incorrect information in their responses compared to the responses provided by physicians. This research shows that although language models such as Med-PaLM and Med-PalM 2 can be useful in some situations, they are not perfect and have limitations that human clinicians do not have.

However, this recent study found that Med-PaLM 2 performed as well as real doctors in almost all areas except for a few specific measures.

The WSJ reports that during the Med-PaLM 2 trial, customers have control over their data, meaning they can decide who can see it and how it’s used. Additionally, the data is encrypted, making it difficult for anyone (including Google) to access it without proper authorization. Essentially, this means that customers using Med-PaLM 2 can trust that their data will be kept private and secure.

Greg Corrado, Google’s chief research officer, told the WSJ that, “Med-PaLM 2 (an artificial intelligence tool designed to help doctors) is still in the early stages of development.” Corrado said that while he would not personally use the tool for his family’s healthcare needs, he believes that Med-PaLM 2 could address areas in healthcare where artificial intelligence could be useful, making them even 10 Equally more effective. In fact, according to Greg Corrado, Med-PaLM 2 has the potential to be a valuable tool in healthcare once fully developed.