@FinanceYF5: AI enables students to score nearly full marks on their assignments, yet their issues are collectively exposed during t…

X AI KOLs Following News
ai-in-education stanford academic-integrity students teaching-assistants

Summary

AI帮助学生在作业中拿到接近满分，但在线下期末考试中却集体暴露问题。一位斯坦福教授表示，学生用AI完成作业后不再参加答疑，却无法通过线下期末考试，因此院系重新定义助教职责，取消传统答疑，要求学生一对一向助教讲解每份编程作业以证明真正理解代码。

AI enables students to score nearly full marks on their assignments, yet their issues are collectively exposed during the in-person final exams. A Stanford professor revealed: After completing homework with the help of AI, students no longer attend office hours, yet they fail to pass the offline final exams. The department has therefore redefined the responsibilities of teaching assistants: Cancel traditional office hours and require students to explain each programming assignment one-on-one with the TA, proving they truly understand the code.
Original Article

Similar Articles

@Gorden_Sun: https://x.com/Gorden_Sun/status/2066919099016630286

X AI KOLs Following

A long-term study involving 26,000 Chinese middle and high school students found that after students independently used AI, homework performance improved by 18%, but closed-book exam scores dropped by 20% within six months. Zhongkao and Gaokao scores dropped by 24% and 18% respectively, and 81% of students used AI to complete their homework.

@Fenng: AI little princes are shocked again... This problem has always existed. Using question banks like '作业帮' directly to search for answers (such tools have already evolved into AI), handing over thinking to tools, the inevitable result is a decline in one's own abilities. What's surprising about that?

X AI KOLs Timeline

This tweet discusses a study on students using AI to assist with homework, pointing out that while short-term grades improve, long-term exam scores decline, prompting reflection on the role of AI in education.

@FinanceYF5: Career advice in the AI era from someone who has worked at Scale, OpenAI, and Google 1/ AI can solve all problems with standard answers School exams, Leetcode, system design — these are becoming AI's home turf What's truly valuable is the ability that can't be written as a loss function

X AI KOLs Following

A practitioner who has worked at Scale, OpenAI, and Google shares career advice for the AI era: AI is good at solving problems with standard answers (e.g., exams, Leetcode), while the truly scarce abilities are those that cannot be defined by a loss function.