Google announces progress on AI for various medical applications, utilizing both the existing PaLM 2 model and the new Gemini model.
PaLM 2:
A year ago, Google introduced the PaLM 2 language model of 2023 and fine-tuned it for medical purposes under the name Med-PaLM 2. It has since been developed into a series of models known as MedLM, available for renting on Google Cloud’s Vertex AI service.
Recently, Google released MedLM for Chest X-ray to assist in interpreting chest X-ray films (such as lungs and heart). Customers can now test this limited-release version.
Gemini:
The latest model, Gemini, has also been deployed for medical use. Google describes Gemini as a multi-modal model that supports various input types including text, sound, images, and videos, making it suitable for medical tasks with diverse input types right from the start.
Google reported successful testing of Gemini on a similar exam to the U.S. Medical Licensing Exam (USMLE) with a score of 91.1%, indicating significant progress. They are also developing Gemini to interpret X-ray films and even brain CT scan images.
On another front, Google is partnering with its subsidiary, Fitbit, to develop Gemini into a Personal Health Large Language Model for use in Fitbit and Pixel devices. This model will enable devices to read and analyze our health data instantly. Google is currently testing the model’s ability to analyze sleep quality data with promising results.
Source – Google
TLDR: Google is advancing AI in medicine with models like MedLM and Gemini, tailored for diverse medical applications, achieving promising results in tests and collaborations with Fitbit.
Leave a Comment