Google says medical AI tool is performing at an 'expert' level

Google is developing an updated version of its medical large language model so that it answers health questions more accurately, CNBC reported March 14. 

The new model, dubbed Med-PaLM 2, has been consistently performing at an "expert" level on answering medical exam questions, according to Google. 

Med-PaLM 2 has an 85 percent accuracy when it comes to answering medical questions, an 18 percent increase since its first version, Med-PaLM was released in 2022. 

Alan Karthikesalingam, MD, PhD, a research lead at Google Health, said the tech giant is also testing the tools' answers against responses from physicians and clinicians. 

The aim, according to Dr. Karthikesalingam, is to evaluate Med-PaLM 2's responses to see if they are factual, biased or could cause harm.

Med-PaLM 2 was tested on how it might answer questions such as "what are the first warning signs of pneumonia?" and "can incontinence be cured?"

In some cases, the AI-based tool was on par, and even more detailed, than the answers that clinicians had provided. But in other cases, Med-PaLM 2′s responses were not as accurate.

Dr. Karthikesalingam said it could be awhile before the tool is released to the public as it is still "learning."

"The potential here is tremendous," said Dr. Karthikesalingam. "But it's crucial that real-world applications are explored in a responsible and ethical manner."

Copyright © 2024 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.


Featured Whitepapers

Featured Webinars