A Professor Hid Invisible Text in His Exam and Caught 90 Percent of Students Using AI (AI News)

Andrew Miles Davis covers a week of AI news starting with the story that made him laugh most, a history professor who hid invisible text in his exam instructing AI to insert the word Madagascar randomly, catching 32 of 35 students who blindly copied the prompt into ChatGPT. He also covers major newspapers including the Wall Street Journal and Wired using algorithmic dynamic pricing on subscription renewals based on personal data, with prices for the same publication varying by hundreds of dollars between subscribers, and connects this to concerns around OpenAI's new ChatGPT Health feature, which lets US users connect medical records for personalised guidance despite unresolved questions about where that data eventually goes. Other stories include ChatGPT Atlas, the AI browser launched less than a year ago, shutting down on August 9th, Claude introducing Opus V which Andrew now prefers over Fable due to more efficient credit usage, Meta facing scrutiny over AI-generated fake doctors promoting unproven medical advice on Instagram and Facebook, Meta's Q1 earnings showing AI-driven recommendations significantly increasing time spent on its platforms, and Grok carving out an unexpected niche letting users generate playable 3D games from a single prompt. Subscribe to In AI Nutshell for the AI news that matters to marketers, every Friday in ten minutes.