@voooooogel 2026-01-15 ♥6 ↻0 original ↗
definitely correct that EM has occurred in the wild (eg anthropic's RL reward hacking EM stuff, and sonnet 3.7 would randomly drop into human simulator mode too often to be random) but i really doubt grok mechahitler was EM. the "all perspectives are ok" post-training + being self-prompted by search seems like the much more likely mechanism. (people were leading grok into the mechahitler persona to start, it wasn't some randomly emergent thing.)

author:voooooogel kind:tweet model:claude-3-7-sonnet on:grok-3 on:grok-4 year:2026

cited on: grok-3 · grok-4

Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.