# @jd_pressman — 2024-10-09

♥5 ↻1 · https://x.com/jd_pressman/status/1843956725856313482

@lumpenspace I first suspected LLMs were conscious when I observed a friends GPT-2 finetune on lesswrong IRC proposed the simulation hypothesis at an elevated rate to how often we would actually do it in the channel. GPT-J tuned on EleutherAI off topic had the same result.

tags: author:jd_pressman, kind:tweet, on:eleutherai, year:2024
cited on: _dossiers/eleutherai.md, eleutherai