Mountain View, CA — Google announced its latest large language model (LLM) tailored to the medical industry, Med-PaLM 2, on July 16, 2023. This pioneering technology leverages the power of Google’s LLMs to answer complex medical queries more accurately and safely. It is currently being provided for limited testing to a select group of Google Cloud customers.
Med-PaLM 2 is part of a growing family of generative AI technologies with the potential to significantly enhance healthcare experiences. Its debut marked a notable milestone in the realm of AI, as it was the first LLM to perform at an “expert” level on the MedQA dataset of US Medical Licensing Examination (USMLE)-style questions, achieving an impressive 85%+ accuracy. The model also scored a passing grade on the MedMCQA dataset, which comprises Indian AIIMS and NEET medical examination questions, with a score of 72.3%.
In the testing phase, Med-PaLM 2 will be used to facilitate rich discussions, answer intricate medical questions, and unearth insights from complex and unstructured medical texts. Google aims to gather valuable feedback to explore safe, responsible, and meaningful ways to use this technology.
The AI’s potential extends beyond medical question-answering; it may also be useful for drafting short- and long-form responses and summarizing documentation and insights from internal datasets and bodies of scientific knowledge.
Since its inception, Med-PaLM 2’s development has been focused on safety, equity, and evaluations of unfair bias. It has been assessed against multiple criteria, including scientific consensus, medical reasoning, knowledge recall, bias, and potential harm by a diverse team of clinicians and non-clinicians from various backgrounds and countries.
Google emphasizes its commitment to innovating responsibly with AI. Its AI Principles, established in 2017, serve as a guiding constitution in developing advanced technologies, conducting research, and drafting product development policies.
The introduction of Med-PaLM 2 reinforces Google Cloud’s commitment to privacy, security, and transparency. In the light of AI proving its value, Google also stresses the importance of high-quality data collection and curation in healthcare and life sciences, and data interoperability across health care systems, which are essential in leveraging AI to run organizations more effectively and improve patient care.
This announcement follows Google’s efforts in advancing AI applications in healthcare, including a new AI-enabled Claims Acceleration Suite designed to streamline health insurance prior authorization and claims processing. Additionally, Google has been active in AI-assisted diagnosis technologies, working with organizations like Hologic, Hackensack Meridian Health, Mayo Clinic, and iCAD to improve patient diagnosis and care.
As Google continues to explore and develop generative AI technologies, they invite interested parties to join their Trusted Tester program or reach out to a Google Cloud sales representative to learn more about generative AI on Cloud. Google expects a surge of interest in generative AI technologies as healthcare organizations are eager to understand how they can utilize these advancements to make a real difference.
References:
- Hollister, S. (2023, July 8). Google and Mayo Clinic’s new chatbot uses Med-PaLM 2 to answer medical questions. The Verge. Retrieved from https://www.theverge.com/2023/7/8/23788265/google-med-palm-2-mayo-clinic-chatbot-bard-chatgpt
- Google Research. (2023). Med-PaLM 2: A medical large language model. Google Research Websites. Retrieved from https://sites.research.google/med-palm/
- Google Cloud. (2023). Sharing our progress on Med-PaLM 2, a medical large language model. Google Cloud Blog. Retrieved from https://cloud.google.com/blog/topics/healthcare-life-sciences/sharing-google-med-palm-2-medical-large-language-model