Google DeepMind developing AI model to assist doctors in diagnosing patients
In recent years, there has been a rapid advancement in artificial intelligence (AI). As a result, major tech companies like Adobe, Meta, and Microsoft have embraced this technology. AI has found various applications, such as ADAS for easier driving and AI chatbots that generate content. Now, Google DeepMind researchers are utilizing AI in the field of medical science. Their latest development involves an AI assistant that aids doctors in diagnosing patients.
In a research paper published on arXiv, Google DeepMind researchers have developed a new artificial intelligence model called Articulate Medical Intelligence Explorer or AMIE.
What is AMIE?
Google researchers say AMIE is a research AI system based on the Large Language Model (LLM) and optimized for diagnostic reasoning and conversations. To scale AMIE’s knowledge and capabilities to different medical conditions and situations, the researchers designed a self-game-based simulated learning environment with automated feedback mechanisms for diagnostic medical dialogue in a virtual care environment.
It is trained on real-world datasets that include medical reasoning, medical summaries, and real clinical conversations. AMIE aims to improve medical diagnostics by improving the quality of medical conversations. It can ask clarifying questions to help diagnose better, while still maintaining empathy.
In an evaluation study conducted by Google, AMIE showed better diagnostic accuracy but also better conversation quality compared to PCP (Primary Care Physicians). Google says, “AMIE had greater diagnostic accuracy and superior performance on 28 of 32 axes from the specialist perspective and on 24 of 26 axes from the patient perspective.”
Can it replace doctors?
No, AMIE is still only a research project. Google says: “AMIE is our exploration into the ‘art of the possible,’ a research-only system that allows us to safely explore a vision of a future in which AI systems might better align with the capabilities of the skilled clinicians entrusted with our care. It’s early experimental work only, not a product, and it there are several limitations that we believe merit further rigorous and extensive scientific research to envision a future in which conversational, empathic, and diagnostic AI systems can become safe, useful, and accessible.”
