Decade of improv shows: early glitchy AI was often funnier on stage than modern models
korymath · x · 2026-10-28
The paper "In and Out of the Circle of Expectations," published in Studies in American Humor, draws on a decade of live Improbotics shows and finds a funny paradox: early glitchy generative models were often funnier onstage than modern AI, for various reasons.
A companion post adds that generating real human laughter remains AGI-hard — about 80% of laughter comes from spontaneous conversation, while most AI humor research focuses on written jokes.
Related event: Making Humans Laugh Remains an AGI-Hard Problem for AI, Study Finds(2 posts)→
More from Fun
- Claude reviews its own plan, admits "actually me neither lol" — teropa · 2026-10-28
- A daily Waymo commute to SF: empty driver's seat, steering wheel turning itself — reed · 2026-10-28
- Developer: the more likes a tweet gets, the smaller the human share becomes — tallinzen · 2026-10-28
- Personal AI agent trains a chess transformer on dual RTX 4090 and battles Sunfish — andreisavu · 2026-10-28
- AI meme: vibecoded 'slop' bragging it outperforms ghostty gets vaporized — generativist · 2026-10-28
- Opus 5.5 ultracode workflow with 9 subagents builds a Three.js scene — nptacek · 2026-10-28