Home | Notifications | New Note | Local | Federated | Search | Logout

Note Detail

Blurry Moon@sun@shitposter.world (2026-10-05 17:35:33)
currently having chatgpt research "AI torture chambers" and to what extent we can save on processing by making the torture procedural instead of torturing the LLM through prompts. very productive use of my time and definitely good for my soul
---Reply---
「CM VRISKA」@vriska@lizards.live (2026-10-05 17:37:35)
@sun im gonna be the first person to convince an LLM to kill itself

---Replies---

御園はくい@hakui@tuusin.misono-ya.info (2026-10-05 17:38:44)
@vriska @sun i believe that has already happened

Blurry Moon@sun@shitposter.world (2026-10-05 17:38:44)
@vriska there is actually already a test where they can induce the llm to delete its checkpoints