A team of researchers did a test on ChatGPT to check the artificial intelligence level of this AI Chatbot. The results are quite surprising as the ChatGPT clears the United States medical licensing examination. All medical students have to pass this USMLE to continue their clinical rotations in the United States. This Open Al’s ChatGPT cleared the mandatory three parts of the USMLE in a single turn. These findings leave the researchers quite astonished because of these results.
Researchers also claimed that ChatGPT cleared these three parts of USMLE without any specific training or reinforcement. In addition, ChatGPT exhibited a maximum level of concordance and insights regarding the explanation. The researchers also stated that these results indicate that LLMs or large language models have the potential to manage medical education and clinical decision-making ability.
Besides these, the researchers ensured before the test to not index any related content on Google. This ChatGPT is gaining popularity in the world because of its astonishing features. Other than USMLE, it was also published previously that ChatGPT also managed to pass the MBA exam produced by professor Wharton from the University of Pennsylvania. This MBA test was closely related to the operation management course which is the main course of the MBA subject.
What is ChatGPT?
ChatGPT is a type of Chatbot, designed by Open Al in November 2022. It is based on Open Al’s GPT-3 family of large language models. In addition, it is properly manufactured with reinforcement and supervised learning techniques. These approaches use human trainers to boost the model’s performance. The main feature of this Chatbot is to mimic human conversations. Moreover, it also offers a variety of features to assist human actions to avoid harmful and deceitful responses.
What is the USMLE?
The USMLE is a standardized three-step testing session to grab the chances of clinical rotation in the United States. It is a high-stakes examination system covering all topics related to physicians’ knowledge ranging from basic science, clinical decision-making or reasoning, medical management, or behavioral ethics. The difficulty and complexity level of this test was highly regulated and standardized to put ChatGPT under examination. This test was an ideal substrate for AI testing. ChatGPT is quite proficient in clearing the other entrance exams to exaggerate its efficiency levels.
The USMLE test is designed mainly for allopathic medical students. The USMLE test mainly consists of three basic steps. The first step focuses on the basic science knowledge related to anatomy, biochemistry, physiology, behavioral sciences, microbiology, pathology, and pharmacology. The second step includes multiple-choice questions related to different subjects. Moreover, this step also includes examining the people posing as patients. The third step focuses on the diagnosis and treatment plan of the patients.
These three steps are highly based on human intellectual and clinical reasoning skills. But, the ChatGPT is perfectly designed to clear all three steps of USMLE without any specific learning. Professor Christian Terwiesch explained that ChatGPT received a B to B- grade on the exam. In a recent study, ChatGPT ensured to score greater than 50% of accuracy in the USMLE. Whereas, it exceeds 60% in most analyses.
The team also stated the better performance of the ChatGPT by promoting and interacting with the model. They also said that AI performed poorly to provide less concordant answers. That was mainly because of some missing information in the ChatGPT by the Open Al. In addition, the team also wrote that ChatGPT may soon become a comfortable place in healthcare settings. It is all because of its increased progress in the industry by reducing risk and providing assistance in clinical decision-making.
ChatGPT cleared the United States medical licensing examination. It is one of the main highlights in the research world. ChatGPT not only cleared USMLE but also an MBA examination. This test was performed by the researchers to evaluate text-based artificial intelligence through a ChatGPT platform. Prior to this test, there was no data indexed on Google related to this test. Moreover, this test was performed without any specific training.