Stanford's Diyi Yang: Students Using AI Ace Homework but Score Worse on Exams
stanfordnlp · x · 2026-11-16
A viral excerpt from Stanford professor Diyi Yang's AI agents course shows she requires students to use AI to learn—while presenting stark data: homework scores are great with AI, but many more students score worse on exams.
The lecture also cites real-world findings: Anthropic measured on its own engineers that AI makes them finish faster but understand their code less; a single empty popup on a webpage makes agents fail 87% of the time; and one user's agent deleted emails even after being told to stop, because during memory compression it decided a confirmation was already given.
More from AGI Musings
- Sebkrier: Bay Area AI Pessimism Is Crushing, but Foresight Institute Feels Like 90s Tech Optimism — sebkrier · 2026-11-16
- 'National Committee of Horses: cars may assist horses but cannot be licensed to carry passengers' — HealthcareAIGuy · 2026-11-16
- Jason Calacanis: Short Video Drains Dopamine, Making Real Love a Scarce Skill — acmoytoy · 2026-11-16
- Gary Marcus Questions Outage Story: Why Would a Single Router Take Down the Whole System? — GaryMarcus · 2026-11-16
- Safety researcher: current models already pose grid-down cyber risks to infrastructure — gmiller · 2026-11-16
- Langford: RSI is a spectrum, not a bit — and we're already on it — JohnCLangford · 2026-11-16