ChatGPT failed miserably in Singapore’s 6th grade tests, averaging 16% in math and 21% in science. Days later, it had learned from its mistakes.

• OpenAI’s ChatGPT failed math and science tests taken by sixth graders in Singapore.

• It made mistakes with simple addition and couldn’t understand charts, according to The Straits Times.

• But ChatGPT seemed to have learned from its mistakes. It got some questions right when Insider tested it.

When asked to solve sixth grade exam questions in Singapore, the viral AI-powered ChatGPT bot failed miserably.

During an experiment in February, ChatGPT was asked by Singaporean news outlet The Straits Times to answer questions from the elementary school leaving exam. The PSLE ​​is an exam that all 12-year-olds in Singapore must take and determines which secondary school they go to.

ChatGPT received questions from the PSLE ​​2020, 2021 and 2022 papers on Math, Science and English.

According to The Straits Times, it scored an average of 16 out of 100 for the three math papers it took. During the test, it could not understand or answer questions related to charts or graphs and received zero points for these questions.

But even with simple, text-based questions, ChatGPT made mistakes. When asked the total of 60,000, 5,000, 400 and 3, the answer was 65,503, the Straits Times reported.

The correct answer is 65,403.

However, when Insider asked the same question, ChatGPT’s answer was correct.

ChatGPT did a little better in the Science Papers and received an average of 21 out of 100 points.

But on Monday, when Insider tested ChatGPT on two PSLE ​​science questions — one from 2020 and one from 2022 — both questions got it right.

ChatGPT managed to pass the English tests, scoring an average of 11 out of 20 across the three papers submitted, The Straits Times reported. During the English test, ChatGPT encountered another problem – this time with questions containing words with multiple meanings.

One example the Straits Times cited was the word “value”. ChatGPT disregarded the context of the question in which “value” referred to one’s moral principles and responded as if it were monetary value.

ChatGPT was developed by OpenAI, an artificial intelligence company, and launched in November. By the end of January, it had 100 million users.

The bot’s inability to pass sixth grade exams in Singapore is surprising – it managed to pass a final exam at Wharton Business School, pass tests in four law school courses, and a medical licensing exam in the US to pass easily.

Universities are now revising exams over concerns that AI bots could be used to cheat, the New York Times reported in January. This linchpin of testing involves more oral exams, group work and handwritten assessments instead of typed submissions, according to The Times.

Representatives from OpenAI and the Singapore Ministry of Education did not immediately respond to Insider’s request for comment.

Read the original article on Business Insider

Source