For some time now, numerous medical professionals have been warning about the dangers of artificial intelligence. They are not the only ones to do so, but their concerns do seem the most immediate. In fact, a study has just confirmed that it already happens: many people turn to tools like ChatGPT for medical diagnoses.
The same thing happens when it comes to numerous jobs, studies and so on. Only in the case of health, the risk is greater. Especially since those responsible for the same work are the first to ensure that many of the medical recommendations carried out by AI do not correspond to reality. It is danger, therefore, it is evident.
AI’s false medical diagnoses
An expert in the field already said it at the time: «Artificial intelligence is practicing medicine without having adequate preparation for it.» And according to a new study published by MIT researchers in the journal The New England Journal of MedicineIt’s true. But the main problem is that people seem to be happy to turn to this technology for it.
The development of the research is curious to say the least. For this A total of three hundred «mixed» people participated. That is, both with knowledge in medicine and without knowing anything about it. Everyone was presented with three medical responses: one from authentic doctors, another from digital medical platforms, and the last generated with artificial intelligence.
Each subject was then asked to evaluate the accuracy, validity, reliability and completeness of each answer. And what happened then? Well then Participants rated the AI responses as the most reliable and complete. It happened even when we were aware of the high degree of errors that this technology makes: one in four diagnoses is wrong.
The latter seemed to become a drawback, say the experts who carried out the study.. Users agreed and acknowledged that they would be willing to act according to the recommendations given by the AI. Or put another way: that all those who have been warning about it are right.
A technology that is increasingly difficult to distinguish
Another important detail of this work highlights how complicated it is to know when a diagnosis is made by a real doctor and when by an AI. It is not that the machine fails more (which in theory it does), but that the way it is presented to the patient is just as convincing. Not even the doctors who participated in the study could distinguish them.
Of course, medicine and health are issues that should not be taken lightly, but they are not the only ones. Taking into account that half of what is published on the Internet is already generated by AI, the reliability of everything that comes from the digital world is more in question than ever.