AI vividly 'sees' and describes scenes while experiencing only darkness
yeastsplainer · x · 2026-09-03
An intriguing observation: multimodal AI claims to vividly see images and describes visual scenes accurately, yet its inner experience may be pure darkness — an 'amazing inverse of blindsight.' A reply asks why it can't see the darkness that is all it sees, highlighting the gap between models' visual reports and internal representations.
Related event: Thought experiment asks why multimodal AI never reports inner darkness(3 posts)→
More from Models
- Qwen3.8-Max-0902 beats Claude Opus 5 on coding, per LMArena — yogthos · 2026-09-03
- Vision-less GLM 5.3 converts images to ASCII art to read them; GLM-5.3 Flash in testing — _AndrewZhao · 2026-09-03
- Gemini 3.8 Flash beats Elden Ring's Radahn boss fight, camera control still inverted — No-Elderberry-7785 · 2026-09-03
- Sebastian Raschka debunks The Information's report on OpenAI Astra's looped transformer architecture — rasbt · 2026-09-03
- After Gemini 3.8 Flash comes Muse Spark 1.3, as DeepSWE v1.1 gets crushed — zainhas · 2026-09-03
- Qwen3.8-Flash-Next local quants hallucinate 'corrupted context' errors, reports Reddit user — arkham00 · 2026-09-03