OFweekrobotNet News Lei Feng Net Press: This article is translated from “FierceHealthcare”. The author believes that AI is overhyped and the value it can generate is still quite limited.
“The turmoil surrounding AI and machine learning is stronger than ever, and it’s distracting the medical community,” said two Stanford University researchers. They believe that AI has reached its expected peak, and the medical industry needs to focus on how this cutting-edge technology can actually improve medical care.
Jo from Stanford University’s Department of MedicineNathan H. Chen, MD, and Steven M. Asch, Ph.D., said in the New England Journal of Medicine that “AI has reached its expected peak,” and that “by increasing our understanding of the capabilities and limitations of AI technologies, the healthcare industry can slow down.” And then on to the footsteps of disappointment.”
An extreme example came earlier this month: Venture capitalist Vinod Khosla said AI will soon replace oncologists. “It’s hard to think that oncologists can be valuable when there’s a wealth of data on tumors,” he said at an event hosted by MIT, according to VentureBeat. “They don’t understand what’s going to happen in the future.” Class debates are exciting but useless. Ultimately, the healthcare industry can only truly benefit from collaboration between medical experts and algorithms.
The real value of AI lies in assisting decision-making, Chen and Asch added. Current AI systems often draw conclusions about what patients already know, and if AI is to be of real value in clinical care, the technology must provide predictions that can influence clinical decision-making and improve clinical practice.
“The standard for looking at AI systems must be the standard of care in the real world. For example, doctors will misestimate the positive rate of rare disease screening, and there is a one-third probability of overestimating the patient’s life expectancy. intensive treatment,” the researchers wrote.
Until then, the industry must overcome this mountain: how to make it easier to access valuable patient data in real time. As Chen and Asch pointed out, the shelf life of clinical data used for prediction is about 4 months.
Needless to say, while many companies in the medical world are investing heavily in AI, the practical application of AI is still quite limited.
The Links: 3HNA009326-001 3HAC025466-001