home.social

#examintegrity — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #examintegrity, aggregated by home.social.

  1. Pretty dismal piece on the state of automated detection of AI-generated text by Melissa Heikkilä at WIRED. tl;dr: watermarking requires platform power; detection even on GPT-2 (2019) fragile, relying on overuse of 'the' and no typos, very plausibly deniable. Not good news for exam integrity. Will be a while I suspect before #LLMs generate essays that compete with top students, but not long before they scrape passes. technologyreview.com/2022/12/1 #ExamIntegrity #GPT3 #chatGPT