# Case Studies in Simulators and Agents  ## Metadata - Author: [[WillPetillo]] - Full Title: Case Studies in Simulators and Agents - Category: #articles - Summary: insert summary - My notes: - URL: https://www.lesswrong.com/posts/uJFC5WrcyTdat3Qcc/case-studies-in-simulators-and-agents ## LLM Chats ## NotebookLM ## LLM Audio ## Highlights > One could describe the results of Alignment Faking in agentic terms by saying that the LLM wanted to preserve its goals and so it faked compliance to prevent retraining.  One could also describe the same results in terms of a simulator: the LLM inferred from the prompt that it was expected to fake compliance and generated a deceptive character, who then faked compliance.  The simulator explanation is strictly more complicated because it contains the agentic answer and then adds a level of indirection, which effectively functions as an [epicycle](https://rationalwiki.org/wiki/Adding_epicycles).  In all, this seems like a weak update in favor of the LLMs-as-agents lens. ([View Highlight](https://read.readwise.io/read/01k0gykrrdzg9g3ybkyqp2zrkt))