(ReliableNews.org) – It’s not unusual for a person to receive a misdiagnosis from a doctor. In fact, women regularly speak out about this persistent problem. The NBC show “New Amsterdam” once aired an episode where an iPad was making medical diagnoses after just hearing patients’ symptoms and reviewing other information that was automatically uploaded. What if something similar were to become a reality?
In 2022, Axios reported on a new artificial intelligence (AI) tool called ChatGPT. It was able to do incredible things when generating text. Recently, the tool was part of an experiment carried out by Ansible Health. The startup wants to improve its COPD treatments through AI and other tools.
ChatGPT was tested with all three steps of the US Medical Licensing Examination (USMLE) process. According to the results of the study, the AI program passed all of the exam components without any reinforcement or specialized training. Although ChatGPT didn’t pass by much, it was still quite an accomplishment since human students spend hundreds of hours preparing for the tests.
The study was published online, and some pointed out the paper was written both about ChatGPT and co-authored by ChatGPT.
I love how #ChatGPT is both the “subject” and the author of this study 😅 https://t.co/0EezLoDZ9T pic.twitter.com/20VjK54zBi
— Bardia Khosravi, MD (@khosravi_bardia) January 11, 2023
An AI research expert was shocked to see the program listed as a co-author of the paper.
Here we go. I guess that ship has sailed. Researchers started adding ChatGPT as co-author on their papers 🫣
PS: Dear researchers, you forgot the version number!https://t.co/OpA2HQ4CtU
(thx @3scorciav for sharing this!) https://t.co/Bz0lTE0UHE pic.twitter.com/8xyKwuuraW
— Sebastian Raschka (@rasbt) January 12, 2023
Regardless of whether or not ChatGPT wrote the paper, the human researchers believe the program could one day be used for clinical decisions and educational purposes. Ansible CEO Ming Jack Po, who was once a Google project manager, told Axios in a recent interview that the research team was “pretty amazed at the results.” He said the program was not just answering the questions correctly, but it was also very good at “explaining itself.”
Despite how well the AI program performed, it will likely be a long time before these programs are put to use in clinical settings. In the immediate future, people will still have doctors to diagnose their maladies. Even if AI programs are put into practice down the road, it’s not clear exactly how they would be used.
Copyright 2023, ReliableNews.org
