Google Research created an AI system that outperforms human doctors at diagnosing through conversation.

[Darryl]

Google Research created an AI system that outperforms human doctors at diagnosing through conversation. blog.research.google/2024/0...

The Articulate Medical Intelligence Explorer (AMIE) is built on a Google LLM, and it's been optimized for "diagnostic conversations." Essentially, you give AMIE a patient's medical history as a knowledge base to work with, and it takes the patient through a conversation and diagnoses potential issues with incredible accuracy.

AMIE was tested in a controlled environment against primary care physicians using a randomized, double-blind study. The results showed that AMIE performed on par or better than the doctors in simulated diagnostic conversations across the board.

These results suggest that advanced, fine-tuned AI systems can match or exceed the diagnostic capabilities of even highly trained human physicians!

The study had limitations. The text-only interface used is not representative of most clinical interactions, and AMIE has not been tested in any real-world applications yet. Transitioning from a research prototype to a practical, safe and reliable tool for clinical use will present difficult challenges, so we may not see AI like AMIE helping out in our local hospitals any time soon.

Despite these challenges, the potential of AI like AMIE in revolutionizing patient care is immense. Thanks to your support, Malecare helps to focus AI innovation towards oncology, and, specifically, innovation in service for prostate cancer patients.

Please show your support with a quick and cost free way to build prostate cancer awareness. When you click on the Malecare YouTube "subscribe" button, you tell Google/YouTube that prostate cancer is an important topic for you and probably thousands of others like you. Clicking on "subscribe" literally costs nothing. You are not going to receive emails or notices. You don't have to fill in any forms. All you will be doing is something good. You are raising prostate cancer awareness.

Please take a moment right now to click this link and "subscribe."

youtube.com/c/malecare?sub_...

Thanks.

[Thunderball1]

I have two PhDs: in AI/machine learning and human experimental (ecological) psychology and natural intelligence (first one quit after two years to switch, from the AI one to the natural intel one) and a master's degree in AI and machine learning, and a BSc in computer science.:

DO NOT TRUST AI.

It is from a foreign, supernatural, fake source. The mystics and gnostics have known this for centuries. Even movies are made about it. Beware their source of information.

Trust your self. If you don't know why, look into yourself . AI is only textual information. It has nothing to say beyond that. We are not simple books or intellect. Consider our real connection to the environment through other kinds of patterned energy = information. That is the key. AI is dumbing the doctors down even more and ultimately nobody will know anything. Is that the future we want?

[Graham49]

One problem with doctors is that they don’t always read (or perhaps taken in) all your medical history. I have had this happen several times. Once it almost cost me my life with sepsis. It should be possible for AI systems to overcome this shortfall, as long as the medical history in the database is complete.

[Shams_Vjean]

Something I noticed several years ago when going to a local pharmacy’s Urgent Care clinic was how the NP diagnosing my condition was using a flow chart based program to guide them through the process. GIGO is always a danger with this sort of arrangement; but the more a process is based in a scientific method then, IMHO, the better chance it has of successfully achieving the goals it is designed to fulfill. And properly discerning the best use for any tool is always an important factor in a successful outcome.

I appreciate MALECARE’s efforts to advance the use of AI & LLM technology for the benefit of prostate cancer patients. Aside from religious deification, and some might argue in spite of it, perfection is an illusion. And I think everyone knows what we call the one who graduated last in their class at Medical School; so there’s that too.

[mababa]

Thanks for the link, Darryl. There’s no doubt that AI will be able to outperform our MOs. In many illnesses, it comes down to computing best percentage outcomes from an enormous quantity of data. Computers will be better at compiling options tailored to specific human conditions faster and more accurately than humans. As long as humans retain the ability to overrule the answers, we should be safe. AI can already defeat the best chess grandmasters.

[swwags]

Would love to see you interview the doctors and techs working on developing AI software that can accurately (or at least consistently) measure SUV Max. That's going to be a big one. I understand there are a few companies (Venture startups?), racing to be the first.

[cancerfox]

A few years ago I made an appointment with an allergist for allergy problems. I had to go online and fill out several pages of background information before the appointment. Then when I got at his office I had to spend several more minutes talking to the doctor's assistant, who basically asked me a lot of the same questions that I had already answered on the online form. Finally, the doctor came in the exam room and repeated some of my information back to me, but he got a lot of it wrong and I had to correct him. Then he gave me an order to have some blood work done, and told me to return after the blood work results came in! Blood work for allergies? I would have loved to have been consulting with AI that day instead of this human! I immediately started looking for another allergist. 🦊

[Gl448 in reply to cancerfox]

Blood tests aren’t surprising for allergy complaints.

Not sure what tests were ordered for you, but even a basic CBC has indicators of allergy/parasite activities (EOSINOPHILS).