Marking Exam Done by A.I. - Sixty Symbols
The video documents a live experiment in which the hosts upload a second‑year undergraduate quantum mechanics exam into ChatGPT and ask the model to answer as a student would. They then mark the AI‑generated responses using the official solution key, treating the output exactly as they would a human paper. ChatGPT achieved 71 out of 75 marks, equivalent to a 95% score and a UK 2:1 classification, surpassing a previous study where GPT‑3 earned a 65% (also a 2:1). The model correctly solved most quantitative problems, though it occasionally arrived at the right answer for the wrong reasons or made sign errors that required manual adjustment. The hosts note that the AI’s reasoning is largely pattern‑matching rather than genuine understanding. Notable moments include the quip “Education as we know it may well be dead,” and a candid discussion about the inadequacy of current AI‑detection tools, which often fail to flag AI‑generated text or solutions. The presenters also highlight institutional pressures to shift toward online assessments, raising concerns that such formats could be easily gamed by AI. The experiment underscores the urgent need to rethink assessment design, incorporating safeguards against AI‑assisted cheating while exploring how generative tools might augment learning. Universities must balance the efficiency gains of digital exams with the risk of eroding academic integrity and devaluing genuine mastery of complex subjects like quantum mechanics.
Top Science Advances of 2025 - Roundup Stream
The livestream serves as a rapid, Wikipedia‑sourced roundup of the year’s most notable scientific breakthroughs, spanning astronomy, physics, chemistry, and engineering. Hosted on New Year’s Eve, the presenter walks through dozens of discoveries, offering brief commentary where possible. Among the highlights,...