For a chatbot that cleared Wharton’s MBA, four law school courses, and a US medical licensure exam with ease, it is a bit surprising that it failed primary-level school exams. Seems that OpenAI’s ChatGPT isn’t coming for our jobs any time soon.
A few weeks ago, ChatGPT made deadlines for clearing Wharton’s MBA exam, as well as scoring better than most students in the United States medical exam to secure a practising license.
However, it turns out, that ChatGPT might not be that smart after all. Recently, OpenAI’s chatbot failed Singapore’s 6th-grade maths and science tests by a significant margin. It seems that ChatGPT isn’t coming after our jobs anytime soon after all.
ChatGPT’s abysmal score in a basic primary school examWhen the popular, AI-powered ChatGPT bot was challenged to answer questions from Singapore’s sixth-grade exams, it failed terribly.
During an experiment in February, the Singaporean news organisation The Straits Times requested ChatGPT to answer questions from the Primary School Leaving Test. The PSLE is an exam that all Singapore 12-year-olds must take to select which secondary school they will attend.S
ChatGPT was given questions from the PSLE math, science, and English examinations for 2020, 2021, and 2022.
According to The Straits Times, it received an average of 16 out of 100 points for the three mathematics papers it took. Throughout the test, it was unable to grasp or answer any questions involving diagrams or graphs and received zero points for these problems.
Nevertheless, ChatGPT also made errors with basic text-based queries. When queried for the figure of 60,000, 5,000, 400, and 3, it responded with 65,503, according to The Straits Times.
In science, it did a little better, scoring 21 out of 100. Still, not exactly a passing score.
Did well in English, but still scored below averageAccording to The Straits Times, ChatGPT passed the English examinations with an average of 11 out of 20 marks across the three papers it took. ChatGPT had further issues throughout the English test, this time with questions involving terms with different meanings.
The term “value” was one example given by The Straits Times. ChatGPT ignored the context of the question, in which “value” related to one’s moral convictions, and responded as if it meant monetary worth.
An MBA, a doctor and a lawyer that was stumped by basic primary school questionsThe bot’s failure to pass Singapore’s sixth-grade examinations is remarkable given that it completed a final exam at the Wharton business school, four law school courses, and a US medical licensure exam with ease.
The New York Times reported in January that universities are redesigning exams in response to worries that AI bots may be used to cheat. According to The Times, this shift in testing will include more oral examinations, group projects, and handwritten assessments rather than typed responses.
The bot learnsOne good thing about the bot, that, unfortunately, many human students don’t have, is the bot’s ability to learn from its mistakes. When pointed out that its answers were wrong, the bot apologised and seemed to learn where it made a mistake.
When it was asked the same questions again, after a few days, it answered most of them correctly, but still scored badly on the tests, mainly due to the visual elements of the questions.
Read all the Latest News, Trending News, Cricket News, Bollywood News,India News and Entertainment News here. Follow us on Facebook, Twitter and Instagram.