- Google on Tuesday announced a series of new health initiatives and partnerships at its annual event called “The Check Up.
- The Google Health team shared updates about features coming to search, tools for building health apps, and the latest in AI-powered health research.
- There was a particular buzz around the latest version of Google’s medical large language model called Med-PaLM.
Google on Tuesday, announced new initiatives and partnerships in health during its annual conference named “The The Check Up.” Google Health’s Google Health team shared updates on new features coming to the search engine, tools for creating health apps, and the most recent developments in technology in health research powered by artificial intelligence.
The company discussed partnerships with new partners to develop AI-assisted ultrasounds, cancer treatments and tuberculosis tests There was also an emphasis on the most recent edition of the medical large-language model, which is dubbed Med-PaLM.
Google launched Med-PaLM for the first time just in the year. It was designed to give high-quality answers to medical inquiries. Med-PaLM is the first AI system that was able to earn an acceptable score, or higher than 60% on multiple-choice questions similar to those found on U.S. medical licensing exams.
The company claimed that the second version of the software, Med-PaLM 2 consistently scored on the “expert” level in Medical exam-related questions. Med-PaLM 2 has surpassed 90% accuracy and has scored 18% more than its previous scores.
Doctor. Alan Karthikesalingam, a researcher in Google Health, said the company is also evaluating Med-PaLM’s responses against those of real clinicians and doctors. He explained that Med-PaLM’s answers are scrutinized for their accuracy as well as bias and the risk of harm.
Karthikesalingam demonstrated controlled examplesalthough the demonstration was not live of how Med-PaLM 2 could address questions such as “what are the initial symptoms in the case of pneumonia?” And “can incontinence be managed?” In some instances, Med-PaLM 2’s answers were comparable with, and sometimes even more specific and precise than the responses that doctors had offered. However, in some instances the responses of Med-PaLM 2 were not as precise.
“You will see through this kind of job that there’s a lot we’re in the process of learning” Karthikesalingam stated during the occasion.
Due to the delicate nature of medical information Karthikesalingam believes it will take some time before technology like this is accessible to consumers on a regular basis. He stressed that it is essential to develop responsibly as well as in a safe and controlled environment.
Google continues to collaborate with experts and researchers on Med-PaLM. Karthikesalingam stated that Google will be sharing more updates on the issue in the near future.
“The potential for this is enormous,” he said, “but it’s essential that the applications in the real world are investigated in a responsible and ethical way.”