OpenAI launched ChatGPT, a long-form question-answering AI that replies to complex questions in a conversational style.
Many users are impressed by its ability to provide human-quality responses, raising the possibility that it could one day disrupt how humans interact with computers and change how information is retrieved.
ChatGPT and its various use cases are still a hot topic in the tech world this year. ChatGPT is being hailed as a game changer in education, and it appears that the chatbot is capable of doing a very good job when it comes to passing exams.
ChatGPT is basically large language model chatbot based on GPT-3.5. It has an amazing ability to interact in the form of a conversational dialogue and provide responses that can appear surprisingly human.
The task of predicting the next word in a series of words is performed by large language models. Reinforcement Learning with Human Feedback (RLHF) is an additional layer of training that uses human feedback to help ChatGPT learn to follow directions and generate human-satisfactory responses.
Since its launch, ChatGPT has crossed its 1 million user mark in December.
According to a new report,ChatGPT has passed several examinations, including law exams from the University of Minnesota courses, parts of the U.S. Medical Licensing exams, and MBA exam from the University of Pennsylvania’s Wharton School of Business.
let’s discuss in detail about the exams passed by ChatGPT
Wharton’s MBA exam
ChatGPT took an Operations Management exam assigned by Christian Terwiesch, a professor at Wharton Business School. After grading the paper, Terwiesch revealed that the model scored between B and B- (B minus), which is a respectable score for a student in the course.
Terwiesch went on to say that the answers were not only correct, but also well-explained. However, the model made errors during relatively simple calculations at an elementary level.
The current AI model does not appear to be capable of solving advanced mathematics. He went on to say that the AI has the ability to change its response when given hints about the correct answer.
United States Medical Licensing Examination (USMLE)
The United States Medical Licensing Examination is widely regarded as one of the most difficult medical exams in the world. It is a three-part exam that takes over four years to complete. The exams include difficult questions ranging from basic science to medical management.
ChatGPT passed all three exams with over 60% marks and an accuracy rate of 50% on the majority of the questions. The researchers were astounded that the bot accomplished this feat without any specialised training.
Multistate Bar Exam (MBE)
ChatGPT achieved a 50% accuracy rate on the Bar Exam’s multiple-choice questions. The MBE is the first of three Bar Exam examinations that qualify a student to practise law.
However, the AI was only tested on the Torts and Evidence sections, which account for two of the seven subjects. Without any prior training or knowledge, the bot answered correctly 50% of the questions. The researchers stated that a large and developing model would perform even better in the future.