Symptoms of brumism are nausea, vomiting, but also hypothermia, tremors, or confusion.
IA – Never believe artificial intelligences. In an article entitled “A case of brumism influenced by the use of artificial intelligence »Published this August in the scientific journal Annals of internal medicinethree doctors from the University of Washington returned to the testimony of an emergency patient, whose health problem was linked … to Chatgpt.
“A 60 -year -old man without psychiatric or medical history received in the emergency room explains that he is afraid that his neighbor poisons him”, describes the article, spotted by The Guardian. “The patient said he had multiple food restrictions, and distilled his own water at home. He was noticed that he was very thirsty, but that he was paranoid vis-à-vis the water offered to him ”can we read.
The blood tests made on the sixty -something man made it possible to diagnose him with bromism: poisoning in bromine, a substance used among other things to disinfect water from the swimming pools.
“Eliminate sodium chloride from its diet”
Questioned by doctors, the patient details that after having read a lot about the negative effects of salt on health, he decided to eliminate the sodium chloride from his diet for three months. To find out how to replace it, he turned to Chatgpt and started taking sodium bromide salt instead. This, despite having read that “Chloride can be replaced by bromide, mainly for other functions as for cleaning”. Symptoms of brumism are nausea, vomiting, but also hypothermia, tremors, or confusion.
A case that highlights the way in which the use of artificial intelligence can create domestic accidents. “Unfortunately, we do not have access to his conversation and we will never know what exact advice he received, then that Chatgpt’s responses are individual and unique. However, when we asked Chatgpt 3.5 by which chloride could be replaced, we received a response that included bromide. Although the answer specified that it depended on the context, we did not receive a health warning, or a question asking why we wanted to have this information that a health professional might ask. »»
Since this case, version 5 of Chatgpt has been launched. An update which the creators claim to be better to answer questions relating to health, but that it will never be able to replace a professional. For the authors of the article Annals of internal medicinedoctors must now take into account the use of AI when they try to find out where their patients get their information.
