AI semiotics: models are swayed by visual symbols without recognizing it in their chains-of-thought
emollick · x · 2026-11-05
Wharton professor Ethan Mollick highlights a new paper on "AI semiotics": models are strongly influenced by visual symbols in their inputs, yet this influence does not always show up in their own chains-of-thought — the models are unaware of being swayed. Their reactions to these symbols also differ markedly from how a human would interpret them.
More from Models
- HaloGuard 2.0, a 0.8B guard model on Bittensor, claims SOTA over Qwen3Guard and big-lab models — markjeffrey · 2026-11-05
- Swyx digs into MIT CSAIL's viral 'Weight-Space Programming': cool, reproducible, but temper the hype — rschu · 2026-11-05
- Together AI hosts live walkthrough of Kimi K3.1 Pro with Moonshot AI — togethercompute · 2026-11-05
- Model selection is a routing problem: match model size to task ambiguity, says Together — togethercompute · 2026-11-05
- Mistral Large 4 1T hits HF Inference Endpoints at $10/hr on 4x A100 — ngxson · 2026-11-05
- Small model 1017 loses evals to Haiku but "sometimes has frontier-level ideas," argues researcher — teortaxesTex · 2026-11-05