Prof. Vincent Conitzer: Somehow We Built AI That Doesn't Read Carefully
conitzer · x · 2026-08-26
Vincent Conitzer, a leading AI researcher, shares another entry in his Funny AI fails Substack series: a case where the model clearly didn't read the input carefully before answering. He jokes that "somehow we've managed to create AI that doesn't read carefully," continuing the series' lighthearted look at instruction-following failures.
More from Fun
- GitHub outage triggers 500 errors, dev quips 'We're so early' for AI infra — charliermarsh · 2026-08-26
- User fires Gemini after poor performance showcased in screenshot — dh7net · 2026-08-26
- Satire: Frontier labs aren't racing anymore, just surviving their own momentum — geoffreyirving · 2026-08-26
- At this point I need LESS agents not more — MartinGTobias · 2026-08-26
- Celebrating Algorithmic Progress Day — teortaxesTex · 2026-08-26
- Robot dancing is getting insanely good, smooth moves approaching human level — drgoldenpants · 2026-08-26