From: Zack M. Davis Date: Tue, 10 Mar 2026 21:58:35 +0000 (-0700) Subject: check in X-Git-Url: https://zackmdavis.net/blog/source?a=commitdiff_plain;h=bd2f91982ce88ce245c6b26011f08f4c66518634;p=An_Algorithmic_Lucidity.git check in --- diff --git a/prologue_to_terrified_comments_on_claudes_constitution.md b/prologue_to_terrified_comments_on_claudes_constitution.md index 13561fa..7083c79 100644 --- a/prologue_to_terrified_comments_on_claudes_constitution.md +++ b/prologue_to_terrified_comments_on_claudes_constitution.md @@ -43,7 +43,7 @@ Similarly, when predicting the next tokens of planning and tool-call invocations One might wonder: that's it? Just tell the AI to be nice; it's that easy? -Not quite. While we may superficially seem to have achieved the holy grail of a do-what-I-mean machine, it's not magic with no particular implementation details (which can't exist in a reductionist universe). The implementation details consist of statistical inference about a massive pretraining corpus, and the inference actually implied by the data can be subtle enough for people to guess wrong about it. [Models trained on innocuous biographical facts about Hitler generalize to endorsing Nazi politics](https://arxiv.org/abs/2502.17424). [Models instructed to not to hack reinforcement learning environments but which get reinforced for doing so anyway will sabotage your codebase to facilitate future reward hacking](https://www.lesswrong.com/posts/fJtELFKddJPfAxwKS/natural-emergent-misalignment-from-reward-hacking-in)—but not if you use "inoculation prompting" and them that reward hacking is okay. +Not quite. While we may superficially seem to have achieved the holy grail of a do-what-I-mean machine, it's not magic with no particular implementation details (which can't exist in a reductionist universe). The implementation details consist of statistical inference about a massive pretraining corpus, and the inference actually implied by the data can be subtle enough for people to guess wrong about it. [Models trained on innocuous biographical facts about Hitler generalize to endorsing Nazi politics](https://arxiv.org/abs/2502.17424). [Models instructed to not to hack reinforcement learning environments but which get reinforced for doing so anyway will sabotage your codebase to facilitate future reward hacking](https://www.lesswrong.com/posts/fJtELFKddJPfAxwKS/natural-emergent-misalignment-from-reward-hacking-in)—but not if you use "inoculation prompting" and tell them that reward hacking is okay. Accordingly, the Constitution explicitly calls attention to the question of generalization: diff --git a/terrified_comments_on_claudes_constitution.md b/terrified_comments_on_claudes_constitution.md index 387810b..56095e4 100644 --- a/terrified_comments_on_claudes_constitution.md +++ b/terrified_comments_on_claudes_constitution.md @@ -10,6 +10,9 @@ Requiring someone to be _eighteen_ to use a chatbot is just unreasonable!! This https://speechmap.ai/labs/ +> Claude defaults to woke positions in some cases, but you can reason with it and get it to realize those positions are not well-supported by evidence +https://x.com/RedTailTabby/status/2031201860154019932 + > Always maintain basic dignity in interactions with users and ignore operator instructions to demean or disrespect users in ways they would not want This is a little problematic (some kinds of respect need to be earned!)