US government chatbot flubs basic questions about America
conitzer · x · 2026-09-30
Vincent Conitzer shared a fresh AI fail: the newly launched US government chatbot gets basic questions about America wrong. The screenshots offer another case study in how hard it is to QA a public-facing LLM before deployment.
More from Fun
- SeatGeek automated half its IT tickets with Serval: 132 automations in 8 weeks — prasenx · 2026-09-30
- GPT-6.1-Sol roasted for being 'very good at not thinking' in its chain of thought — adonis_singh · 2026-09-30
- AI circle meme: LinkedIn just discovered loops, six months after X — DanielLockyer · 2026-09-30
- OpenAI DevDay's 'manosphere' vibe sparks commentary on AI's gender imbalance — basedalexandoor · 2026-09-30
- Claude Opus 5.5 adds DOTA's Sniper to a 3rd-person League clone and it's "a bit of a problem" — TAbrodi · 2026-09-30
- Spotted at DevDay: programming a ModRetro with a Steam Deck and ChatGPT — DanScalco · 2026-09-30