Google has been testing its medical AI chatbot, Med-PaLM 2, in hospitals, including the Mayo Clinic research hospital since April. The AI tool is designed to provide answers to medical questions and serves as an updated version of PaLM 2. The tool undergoes training using a curated set of medical expert demonstrations to enhance its healthcare conversation capabilities. According to an internal mail, Med-PaLM 2 can be particularly helpful in countries with limited access to doctors.
Ongoing improvements planned to refine Med-PaLM 2 and create a more accurate chatbot
However, a research paper published by Google in May shed light on some accuracy issues with the chatbot. Study shows Med-PaLM 2 had more inaccuracies and irrelevant information compared to answers from other physicians. Improvements are necessary to enhance the chatbot’s performance and reliability in providing accurate medical information.
Despite accuracy issues, the chatbot excelled in reasoning, providing consensus-supported answers, and demonstrating comprehension. Inaccuracies are not unique to Med-PaLM 2; popular chatbots like OpenAI’s ChatGPT and Google’s Bard face similar challenges. The researchers aim to refine the chatbot to create an accurate and meaningful chatbot through ongoing improvements.
Data encryption ensures data remains inaccessible to Google
Google’s Med-PaLM 2 is an AI-based healthcare system that is still in its early stages. The system is designed to grant users personalized medical advice while giving customers control over their data. The data will undergo encryption to ensure its security and will remain inaccessible to Google. Greg Corrado, Google’s senior research director, believes that Med-PaLM 2 has the potential to expand the benefits of AI in healthcare significantly, though he wouldn’t personally rely on it for his family’s healthcare.
