The ChatGPT bot, which was driven by artificial intelligence (AI) and became viral, failed when tested on questions from Singapore's sixth-grade exams.
OpenAI, a startup specializing in AI development, developed and released ChatGPT in November 2022. Earlier in January, it had reached 100 million active users.
Failed Exam
In February, The Straits Times conducted an experiment by asking ChatGPT to answer questions from the Primary School Leaving Exam (PSLE). Results were posted on Monday, Feb. 20.
Notably, all Singaporean children aged 12 are required to take the test to choose their secondary school placement.
Questions from the PSLE math, science, and English tests in 2020, 2021, and 2022 were fed into ChatGPT.
The Straits Times said that ChatGPT scored 11 out of 20 in its three English tests.
Meanwhile, the average result on the three math exams it took was 16. Any questions on the exam that required an interpretation of a diagram or graph were disregarded entirely due to the bot's inability to comprehend them.
Moreover, ChatGPT erred on straightforward textual inquiries. According to the report, the AI chatbot gave the wrong number of 65,503 when asked to add 60,000, 5000, 4400, and 3.
Close enough, yet, it is incorrect. The right sum is 65,403.
Redeemed Scores
Days later, Business Insider also asked ChatGPT a similar set of questions. Surprisingly, it got the right response that time.
On average, ChatGPT scored 21 out of 100 on scientific papers, which is an improvement in its overall performance. Business Insider tested ChatGPT on two PSLE science questions from 2020 and 2022 on Monday, and it correctly answered both.
It is stated that ChatGPT passed the English exams with an average score of 11 out of 20 across the three exams it took. But even though ChatGPT had improved, it still struggled with the English exam, especially with questions that used terms with numerous meanings.
The term "value" was used as an illustration. ChatGPT replied as if "value" related to monetary worth when it clearly referred to the questioner's moral convictions.
Effects on Education
Although the bot did pass a final exam at the Wharton Business School, assessments in four law school courses, and the US medical licensure exam. Therefore, its failure to pass Singapore's sixth-grade exams comes as a surprise.
Concerned that AI bots may be exploited for cheating, universities are allegedly revising tests, as reported previously. Due to this shift, students should expect to have more oral examinations, participate in group projects, and submit more handwritten work for evaluation.
More school administrators are addressing the problem by changing their curriculum and blocking ChatGPT. New York, Los Angeles, Seattle, and Fairfax County public schools have prohibited the usage of AI tech.
Meanwhile, OpenAI CEO Sam Altman argues that schools should stop worrying about it.
