Go Tokyo Your Experience
In recent years, with the development of artificial intelligence tools such as ChatGPT, more and more students have begun to use such software to complete their homework, which has posed a considerable challenge for teachers in schools. Although AI technology can generate logically coherent texts, this does not mean that the content is necessarily correct.
Recently, at a private junior high school in Tokyo, Japan, half of the 250 first-year students made the same mistake in their physics and chemistry assignments. After an investigation, the teacher discovered that the source of the problem was actually AI! Let us take a closer look at this interesting incident.
According to reports: “Among the 250 students in the first year of junior high school, over half made the same mistake in their physics and chemistry homework… This left the teacher extremely puzzled, who then discovered that AI incorrectly provided the answers.”
The phenomenon of half of the students in this private junior high school in Tokyo making the same mistake in their physics and chemistry homework raised suspicions among the teachers.
During the correction process, the physics and chemistry teacher discovered a clue: as many as 60 to 70 percent of students in each class used AI to complete their homework. Particularly for the question, “What is the role of amylase in saliva?”
In this question, AI provided an incorrect answer. Therefore, students who used AI to solve the problem all gave the same incorrect answer.
This teacher shared their findings online: “While grading assignments, I noticed that several students included the incorrect description ‘amylase can break down starch in food, making it easier for the stomach to digest.’ In fact, the stomach is not involved in the breakdown of sugars, right? It was through investigation that I found these students’ answers closely matched Google’s AI.”
In response to the question “What role does amylase play in saliva?”, many students answered: “Amylase can break down starch in food, making it easier for the stomach to digest.”
However, the teacher pointed out that if one carefully examines the textbook, it can be found that the organs for digesting starch should be “the mouth and the duodenum,” rather than “the stomach.”
The teacher discovered that the problem originated from incorrect answers provided by Google’s AI, which led students to copy and paste erroneous information. Further investigation revealed that this erroneous information came from the official description on the website of the famous Japanese mayonnaise brand “Kewpie.”
The teacher speculated that it was because the incorrect information on the “Kewpie” official website was recognized by AI and used as an answer, which led to students submitting incorrect assignments uniformly.
The Kewpie official website stated in its explanation of saliva’s functions: saliva aids in food digestion, oral hygiene, and promotes swallowing. However, this incorrect description misled many users who relied on AI.
This event has sparked a heated discussion about the potential dangers of AI after media coverage. Many are concerned that if AI releases incorrect sensitive topic information without strict review, it could mislead those seeking information and lead to negative consequences.
This teacher has suggested to the Kewpie official website to correct its description: “While I’m not sure if they will pay attention to me, I still choose to contact Kewpie in the hope that they can correct the content on their official site.”
Subsequently, Kewpie corrected its description, removing the part about the “stomach” and keeping the description as “Amylase can break down starch in food, making it easier to digest.”
At the same time, the hard work of the relevant teachers in conducting investigations and facilitating corrections is commendable. This incident reminds us that while AI is powerful, it is not infallible and should be used with caution.
