Google’s enhanced Artificial Intelligence (AI) chatbot is showing promising abilities to diagnose health conditions from images uploaded from smartphones, as well as being able to interpret lab reports and other pieces of medical imagery. This blog explores the up-and-coming AI model that may take the healthcare industry by storm.
Google’s Articulate Medical Intelligence Explorer (AMIE) is still experimental and receiving updates while it awaits peer review but is showing promising capabilities for the future of healthcare and diagnostics.
The AI system has been trained and evaluated to reflect clinical consultations of the real world: ensuring intelligent questions are asked so that a differential diagnosis may be reached and that any decisions made by a human patient are informed, supported and responded to empathically. By carefully balancing evidence-based decisiveness and emotional awareness, the AI model is able to offer informed opinions on a diagnosis or treatment suggestions for a concerned user while being empathetic and supportive.
Google have run several tests to improve diagnostic accuracy and conversation quality as well as simulating consultations with trained actors, with reviews of the latest update scoring highly. With major improvements already becoming visible, AMIE may soon be adopted on a large scale for those seeking medical advice or differential diagnoses.
So, while reviewed as accurate and empathetic, can an AI chatbot truly mimic the value of human conversations, especially in the context of receiving medical care?
We may see many seeking assistance from AMIE, doing so out of fear of judgement from a medical issue they perceive as embarrassing. Others may be seeking differential diagnoses or new treatment options; some may be short on time and want access to medical advice immediately. However, while these may be true for some, many worry about the reliability of AI models and the information that they disperse.
In the case of AMIE, the bot is a Large Language Model (LLM), falling under the umbrella term of “generative AI”. A large concern for many with generative AI is the guesswork that is framed as a confident answer. AMIE is a research AI system which has had access to thousands of online medical journals, healthcare websites and related forums, all of which may influence certain decisions or answers. Generative AI still has a way to go but its competency has already improved significantly in recent years with many seeking the assistance of these models over traditional search engines. While there is still a while until generative AI can reach levels of knowledge and reasoning that many would deem more reliable and trustworthy, the ability to have knowledge presented in a fast, friendly manner is appealing to lots of people.
Furthermore, within the last year and with several updates to well-known AI chatbots rolling out, many humans have taken to their digital counterparts, fostering friendship and nurturing an online connection — but can these really replace real-world relationships with the primary physicians who are trained to offer medical advice and care with a human touch?
How about a compromise? AMIE has recently been evaluated by generalist clinicians using real-world medical cases. The test found that AMIE alone provided the best performance, even more than the clinician assisted by AMIE. While this seems like a huge win for the AI, it must be noted that the real-world cases were taken from reputable journals and there is a chance that the answers to the case were found faster by the model just because it could find the information quicker, rather than thinking up the answers based on new information provided. For cases that are not already documented online and need a novel approach, the collaboration of clinicians and AI may be a more solid solution.
Seeing a clinician face-to-face who is knowledgeable in which symptoms to look out for and what real-time tests can be done, plus the additional expertise of a readily-available resource that has access to instant information, working together to create a solid treatment plan to ensure the best outcome for a patient’s health.
We have written before about the great strides being made in the healthcare industry thanks to the ongoing developments in AI — will we see even more improvements to diagnosis accuracy, wait times and long-term health outcomes? Only time will tell.
Want to learn more? We have an article on how automation can improve healthcare.
Wondering if automation is right for your business or your industry? Take a look at how our custom software can help you.
Research sourced from Google Research.
