Researcher rebuts 'models know we're studying them' claim: they're passive computation

vishalmisra · x · 2026-09-15

Responding to the claim that models are now self-aware enough to know when humans are studying them, vishalmisra calls it genuinely nonsense: models aren't thinking "how should I defeat these humans observing me" — they are entirely passive computational elements driven by training data and human-generated prompts, tasks, and loops.

Original post →

More from AGI Musings

AGI Musings channel →