ChatGPT Fails To Perform In This Exam Too, Students Beat AI Tool By 30% Margin

CHATGPT has been the blue-eyed AI bot since its launch last year by OpenAI. It was earlier in the news when it reportedly failed to perform in the Civil Service Examination. And in the latest series, it lagged at an accounting test when put head-to-head with students at Brigham Young University (BYU), according to PTI.

However, the researcher termed the performance of the AI model as "impressive" in the test. The test aimed to learn how the AI engine will perform in an accounting exam. The findings of the result were published in the journal - Issues in Accounting Education.

Also Read: No Work In Progress For GPT-5 As Of Now, OpenAI CEO Confirms

What was the score?

In comparison to the aggregate students average of 77 per cent, ChatGPT could win 47 per cent accuracy, a Hindustan Times report notes. Plus, in about 11.3 per cent of questions, it was better than the competing students. The AI model performed well in domains like auditing and Accounting Information Systems (AIS).

Further, speaking of accuracy in two choice questions (true/false), ChatGPT had a correction rate of 68.7 per cent. On the other hand, MCQs had an efficiency of 59.5 per cent.

Where did it struggle?

It was lacking in finance, tax and managerial assessments. The bot lagged in short-answer questions with a performance range of 28.7 per cent to 39.1 per cent. Further, the language model gave explanatory responses for many answers, the report states. But the answers turned out to be wrong.

While in some instances, it delivered the correct description but opted for the wrong choice in MCQs. Additionally, ChatGPT invented facts with seemingly-true references. It also messed up calculations such as adding the sum in a subtraction query.

A total of 327 co-authors and 186 academic institutions spanning 14 countries participated in the test. The researchers observed it would be a "game changer that will change the way everyone teaches and learns - for the better," the report cites.

Meanwhile, this is not the first time the language model appeared in high-level tests. According to The Indian Express, it passed examinations like US Medical Licensing and Wharton MBA tests.

Inputs: PTI