Can ChatGPT take your tests? AI tool appears to pass medical licensing, MBA exams in US

The questions given to the bot were from previous examinations. Those answers were not available on the dataset that the model had access to

By Web Desk

  • Follow us on
  • google-news
  • whatsapp
  • telegram

Top Stories

Photo: AP
Photo: AP

Published: Mon 23 Jan 2023, 2:52 PM

The artificial intelligence language model, ChatGPT, has recently been going viral for being able to write sophisticated essays.

The Open AI bot was recently part of a research which concluded that the tool was "comfortably within the passing range" of a US medical licensing exam.

The research was conducted using "376 publicly-available test questions". These were obtained from the June 2022 sample exam release on the official USMLE website.

Researchers also performed random spot checking to ensure that none of the answers, explanations, or related content were on Google before January 1, 2022, as the ChatGPT training dataset had access to it.

The model demonstrated a high level of concordance and insight in its explanations, the researchers concluded.

This could mean that such Artificial Intelligence models could possibly help with medical education and even clinical decision-making, the study added.

Titled 'Performance of ChatGPT on USMLE: Potential for AI-Assisted Medical Education Using Large Language Models', the study has been published on medRxiv.

In another similar study reported by Fortune, the AI model was tested on an MBA-level exam by a professor. The expert said that the model would have received "a B to B- grade on the exam", indicating that it did perform however, it required human intervention.

The researcher also mentioned in his report that the limitation to the model was that it was making Grade 6 level math errors, which could be highly problematic.

Christian Terwiesch, a professor at Wharton and conductor of the research also said that this result has important implications for business school education.

ChatGPT has seen a rise in popularity as experts in several fields have been enthralled by the results generated by it. However, the model sometimes gives out "plausible-sounding but incorrect or nonsensical answers", according to its creators.


More news from Business