A new artificial intelligence program developed by researchers at the University of Florida and NVIDIA can generate doctors’ notes that are very similar to those written by real doctors. The study showed that physicians were only able to tell the correct author of the notes 49% of the time. The researchers trained supercomputers to create these medical records using a new model called GatorTronGPT, which is similar to ChatGPT. The program uses large language models to mimic natural human language, particularly in the medical field. The researchers had to gather large amounts of medical data and train the program to write in clinical language. They used a massive dataset of 277 billion words from the OneFlorida+ Clinical Research Network. The study suggests that AI could be used to replace the manual documentation of medical records, and the researchers are working on a commercial version of the software. The team of researchers from UF and UF Health worked on the study, which was funded by grants from various organizations. At the end of the content, there are two paragraphs that reference patient cases written by different authors, and readers are asked if they can tell if the author was a human or the AI program.
>>Join our Facebook Group be part of community. <<