AI in medicine - an attempt to change the rules.
The American Medical Association AMA calls to examine the limits of responsibility of doctors who use artificial intelligence tools in their practice. This was due to a case where a doctor followed wrong AI instructions. The patient dies and the doctor is found guilty.
It seems logical that the developer of AI systems would be responsible for the consequences of the clues he creates. But, as they say, there are nuances.
Medical systems based on neural networks have been widely used in the United States for more than 20 years. Such systems have never had 100% accuracy, and the developers have always stated that such a system does not exempt the doctor from responsibility and can only be used as a recommendation to reduce medical errors.
And that was fine with doctors, because when there is a possibility of error (for example, when processing X-rays or other unclear images), the ability to reduce the likelihood of error is valuable in itself. In the pre-computer era, clinics used the practice of "double reading", where one image was read independently by two doctors, this reduced the percentage of errors.
Computer systems that indicate problem areas in images also reduce the likelihood of errors. And in the Anglo-Saxon system, where it is customary to sue for wrong diagnoses and a mistake has a price in money, it is easy to calculate economically the effectiveness of reducing an error.
Computer systems were actually cheaper to use than a second doctor, despite their huge cost - it paid for itself, that's all. What has changed now?
Instead of "simple" systems based on neural networks, doctors were offered more complex systems that mimic human intelligence. This simulation is probabilistic in nature, errors cannot be avoided - and the developers, realizing this, refuse to be responsible for the results of their system, as they have refused in the past.
However, the name change itself - a marketing ploy - changes the company's attitude towards the device.