Vincent Conitzer shares an AI Overview chess-piece fail: 'maybe some self-awareness?'
conitzer · x · 2026-09-05
Alignment researcher Vincent Conitzer posted a new entry on his Substack series Funny AI fails: an AI Overview treating a chess piece in an oddly anthropomorphized way, prompting his quip that it might have "some degree of self-awareness," with more cases to follow.
Related event: AI Scholar Conitzer Launches Column Collecting Funny AI Fails(2 posts)→
More from Fun
- "I need Starlink on my flight so I can watch my agents": AI reliance hits daily life — gethackteam · 2026-09-05
- Dev Uses GPT-6 Astra to Build 3D Site Splitting Male Anatomy into 2,234 Pieces — i_dg23 · 2026-09-05
- GPT-6 Astra Renders Hyper-Realistic GTA-Style Singapore Street in Three.js — minchoi · 2026-09-05
- GPT-6 Astra Builds Backrooms Entirely in Blender in ~5 Prompts — minchoi · 2026-09-05
- Tesla influencer clique debate: only Cybercab and Optimus results matter, say critics — JOBhakdi · 2026-09-05
- Dev Uses 5.6-luna-low Model to Simulate Dumb Users Before Shipping — generativist · 2026-09-05