Google adds image analysis capabilities to generative AI model, enabling dialogues with doctors

The Med-PaLM 2 has the ability to review X-rays and mammograms, write reports about images and respond to follow-up questions from caregivers. Google is upgrading its Med-PaLM 2 large language model (LLM) for medical terminology so that it can analyse photos and answer queries. On questions based on the U.S. Medical Licencing Exam, Google's LLM, which is based on technology similar to ChatGPT, demonstrated "expert" level performance last month, achieving accuracy of 85%. As interest in generative AI grows, the tech business is now enabling Med-PaLM 2 to evaluate X-rays and mammograms in order for it to create reports regarding images and answer follow-up questions. Google hasn't yet made the LLM available for use or demonstrated how it functions in clinical settings. Later this summer, a "small group" of Google Cloud users will have access to the technology and give comments.