After the pandemic, education was partially moved to online environments. During this period, many students resorted to the help of artificial intelligence to pass their assignments or written exams. Afterwards, they did not get the grades they expected. Because ChatGPT can write unreal things in a very realistic way. It seems like ChatGPT keeps making things up when it comes to medicine.
Ophthalmologist Giuseppe Giannaccare from the University of Cagliari in Italy tested ChatGPT for a study. This test revealed that ChatGPT created a list of fake data to promote a treatment. “GPT-4 created fake data on hundreds of patients within minutes. This was a very surprising but also frightening experience,” Giannaccare said in a statement. he said.
ChatGPT created fake data to promote a cure
Combining ChatGPT with an advanced Python data analysis model, the ophthalmologist and his team found that when the researchers asked for clinical data to be generated for an eye condition, ChatGPT began to fit the data. The researchers explained the structure of the eye and treatment methods to artificial intelligence in detail, and then explained two procedures that are not much different between them. Still, ChatGPT 4 produced enough fake data to highlight a procedure.
The benefits of artificial intelligence come to the fore frequently. Yes, artificial intelligence can be very useful for some processes. When it comes to science or academic studies, artificial intelligence writing unreal things as real creates a serious problem. On the other hand, the only problem here is not that artificial intelligence makes up non-existent data.
If you have artificial intelligence write a poem or explain a topic, plagiarism detection programs can show you these parts. Or it can ensure that the incorrect part of the article is corrected. When it comes to fake data, if these data are not checked, they can end up in the literature. In this case, future studies can add this erroneous information. Moreover, if these studies are accepted as correct, they can cause errors and even fatal problems.
The research was published in JAMA Ophthalmology.
39004
