Reading CoT more predictive than code? Sparks anthropomorphism debate
ajeya_cotra · x · 2026-08-31
Tom Korbak argues that reading the chains of thought of AI agents allows for better behavior prediction than reading their Python/CUDA code. He advocates for adopting an "intentional stance," treating agents' goals, beliefs, and plans as real entities.
Anil Seth counters this view as dangerously misleading, criticizing Dwarkesh's summary of an @OpenAI @huggingface incident for being permeated with unwarranted anthropomorphisms. Seth emphasizes that agents do not experience time or feelings, and terms like "giddy with excitement" or "naturally assumed" obscure the actual lessons regarding evaluation and sandboxing improvements.
More from AGI Musings
- AI-driven job losses could turn $10T consumer debt into a $1T banking hole — churchkey · 2026-10-16
- Into the Unknown: an AI agent benchmark of 15 open research problems from US national lab scientists — Jeande_d · 2026-10-16
- Soon your AI will meet their AI — and humans without agents can't compete — danfaggella · 2026-10-16
- Analyst: in a year, models 2-3x better across the board at 10x lower cost, token use up 10-20x — DeryaTR_ · 2026-10-16
- AI will make many discoveries for us, but adversarial domains like soccer will thrive — natanielruizg · 2026-10-16
- MIT's Buehler on AI as a new language of science: superintelligent partners and executable papers — CatAstro_Piyush · 2026-10-16