Google releases MedLM models for healthcare AI applications



summary
Summary

Google has launched MedLM, a set of specialized generative AI models for the healthcare sector.

The first two models are now available to Google Cloud customers in the US through the Vertex AI platform, while certain markets outside the US have access to the tools as a preview.

The MedLM models are based on Med-PaLM 2, the second iteration of Google’s large-scale medical language models. Med-PaLM 2 has improved by 18 % over its predecessor this year, achieving 85 % accuracy, which Google equates to the level of a medical specialist. The two MedLM models differ in size and use: The first is larger and designed for complex tasks, while the second is a medium-sized model best suited for scaling tasks.

The AI tools are designed for use across the healthcare industry, including hospitals, drug development, or patient-facing chatbots. Google cited several organizations that have already tested the MedLM suite and are now putting it into production or expanding their testing.

Ad

Ad

Google has tested MedLM with large companies, Gemini version coming

In a pilot project at HCA Healthcare, MedLM technology was used to augment Augmedix’s technology, which listens to conversations between doctors and patients and automatically translates them into medical notes, Google said.

BenchSci integrated MedLM into its existing AI-based research and development platform to enhance the ASCEND engine’s ability to quickly sift through large volumes of clinical data and identify correlations between diseases and biomarkers. Accenture and Deloitte have also been testing MedLM tools to accelerate AI-powered clinical document processing, according to Google.

Google plans to integrate Gemini-based models into the MedLM offering in the coming months. The models are expected to deliver better results and expand capabilities. Recently, Microsoft showed that GPT-4 is ahead of Google’s Med-PaLM 2 with the right prompt.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top