If you’ve been following us, you know how passionate we are about mathematical models and applications of data science in healthcare. You must then listen to this brand-new episode of Man & Machine hosted by a well-known family physician and AI enthusiast Dr. Ean Bett.
Dr. Bett’s most recent guest was Daniel Satchkov, WellAI’s machine learning lead, who shared his thoughts and offered his vision for the future of AI and digital health.
Check it out: https://open.spotify.com/episode/5khdhpTxiVhuizgt9LXSSi
Here are some of the excerpts from the interview.
Artificial Intelligence has no intelligence, no creativity and learning. It’s just pattern recognition, pattern repetition and pattern behavior. Humans do have intelligence and creativity.
Daniel: The role of WellAI is to help physicians with the repetitive part and allow physicians to do the creative part, the human part.
At WellAI, we’ve built a model that absorbs millions of medical studies. Based on that it can have a back-and-forth with a human being. It doesn’t replace a doctor. But it replaces some initial questioning.
The WellAI system saves 12-15 minutes per intake. But it’s not just time. When a doctor or a nurse starts talking to a patient, there is already the initial readout of questions and answers available. The doctor can now be fresh and do a better job. Doctors do make mistakes. Part of it is from other problems in medicine. It’s far too skewed to pharmaceuticals, for example. But the other problem is when you do mundane things when you have to step out of a pattern you are no longer functioning. Your mind functions with these heuristics. Whereas it’s much better to have a doctor always being “in the zone”, always doing creative things for patients, and the machine doing the repetitive tasks.
Dr. Bett: When I have interacted with the WellAI app, it actually found some things that may be based on my training and medical education I would not have thought about. Because of these heuristics or the mundane nature of the questions we ask we may miss some things for patients individually that could be of great benefit.
My question is: Was it your intent to find some of these lesser likely diagnoses or considerations because that’s really where I think we can do a great service to each other as individuals, as a physician, and as a patient.
Daniel: That’s right. It was a challenging task. Our goal was to go beyond rules-based and to avoid the EHR dataset biases. The model we built addresses mundane tasks but also the newer and rarer diseases that the EHR and the ER datasets would not be able to handle.