Google Research And DeepMind Create AI Medical Chatbot That Can Generate Safe And Helpful Answers!
Introduction
You can also listen to this article on the YouTube podcast below:
Google Research and DeepMind have created an AI medical chatbot that can generate safe and helpful answers to medical questions. Named MedPaLM, the chatbot combines HealthSearchQA, a free-response dataset of medical questions found online with the latest in artificial intelligence and machine learning technology. By leveraging data from professional medical exams, research, and consumer queries, MedPaLM is able to provide accurate and reliable answers to a variety of medical questions. This breakthrough technology could revolutionize the way we access medical information and advice.
What Is MedPaLM?
MedPaLM is a medical chatbot developed by Google Research and DeepMind that can generate safe and helpful answers to medical questions. This AI-powered chatbot uses large language models (LLMs) which are designed to understand queries and generate appropriate responses in plain language. It is verified by using the open-source benchmark MultiMedQA and is also able to respond to health-related questions. Testing into the new open-source Google LLM involved evaluating its performance by studying its responses for factuality, precision, conceivable harm, and bias. In spite of the fact that the new tool from Google and its AI division DeepMind did not produce the same results as a human clinician, the results were a significant improvement over other similar models.
The testing found that it was most notably inferior to human clinicians for incorrect retrieval of information (16.9% MedPaLM vs 3.6% for humans), proof of incorrect reasoning (10.1% vs 2.1%), and inappropriate or incorrect content of responses (18.7% vs 1.4%). But MedPaLM was able to outperform another LLM, Flan-PaLM. A panel of clinical experts concluded that Flan-PaLM’s responses to long-form answers are accurate 62% of the time while 93% of MedPaLM’s responses were deemed accurate. According to a paper published by the AI tool’s researchers, MedPaLM could have a key role to play in clinical applications following some refinement. The paper stated that the Google tool “performs encouragingly but remains inferior to clinicians.”
This is not the first foray into AI-powered healthcare for Google. In 2021, the company decided to partner with Northwestern Medicine in order to investigate whether artificial intelligence can prioritize mammogram reviews with a greater risk of breast cancer.
How Does MedPaLM Work?
A free-response dataset of medical questions found online, HealthSearchQA, was combined with six existing open-question answering datasets to create MedPalm. MedPaLM is able to address both the quiz, including multiple-choice questions, as well as answers posed by both medical professionals and non-professionals.
Testing into the new open-source Google LLM involved evaluating its performance by studying its responses for factuality precision conceivable harm and bias. Although the new tool from Google and its AI division DeepMind lacked the same results as a human clinician the results were a significant improvement on other similar models that have been investigated.
What Are The Benefits Of Using MedPaLM?
MedPaLM is designed to address the problem of incorrect retrieval of information, proof of incorrect reasoning, and inappropriate or incorrect content of responses which other similar models have failed to do. Although the new tool from Google and its AI division DeepMind lacked the same results as a human clinician the results were a significant improvement on other similar models that have been investigated.
Given these advancements, it is clear that MedPaLM has potential applications in clinical settings. With further refinement, this technology could be used to improve accuracy and speed up response times for medical queries, making it easier and more efficient for doctors to provide safe and helpful answers to their patients.
This article was originally published on medium on 19th of January, 2023.
