The research release version of the AI-powered chatbot ChatGPT developed by the San Francisco-based OpenAI has passed several examinations, including law exams from the University of Minnesota courses, parts of the U.S. Medical Licensing exams, and an exam from the University of Pennsylvania’s Wharton School of Business.
(For insights on emerging themes at the intersection of technology, business, and policy, subscribe to our tech newsletter Today’s Cache.)
While the chatbot has passed these exams, it did not necessarily do so with flying colours and sometimes needed human prompts in order to successfully execute the tasks.
In a paper titled ‘Would Chat GPT3 Get a Wharton MBA?’ researcher Christian Terwiesch assessed the chatbot’s performance in an exam that MBA students would generally take.
Terwiesch said that ChatGPT did an “amazing job” when it came to basic questions related to its field, but sometimes made “surprising mistakes” in calculations at the middle school level. ChatGPT was still unable to handle very advanced questions, but could sometimes evolve its responses after receiving hints.
The news comes shortly after OpenAI’s CEO Sam Altman denied rumours that the upcoming ChatGPT version would be many times more powerful than the current version that has been available for the public to test since around December 1, 2022.
While academicians and invigilators have warned of the dangers of students using ChatGPT to send in unoriginal essays or cheat on tests, companies are working on diverse solutions to detect the use of AI-based tools in submitted work.
