Post #2867472
2026-04-27 10:23 UTC
also, an LLM's "explanation" for its own actions is meaningless and there is no reason to assume anything in the explanation is true. is it still deeply alarming that anyone thinks putting "NEVER FUCKING GUESS" in a text file constitutes a security guardrail
Replies (2)
-
@resuna@ohai.social 2026-04-27 23:18
@jcoglan@mastodon.social Everything an LLM produces talking about itself is parody cyberpunk fanfic.
-
@mirabilos@toot.mirbsd.org 2026-04-28 00:05
@jcoglan@mastodon.social “do not hallucinate” 😹