Eriskii

• Eriskii 12 May 2026 23:24 UTC

  1 point

  −2

  in reply to: Simon Lermen’s comment on: An­thropic’s strange fix­a­tion on hyperstition

  Simulator theory is straightforwardly incorrect as models are trained to predict people in pre training instead of stimulating them

  This doesn’t mean anything, this is an argument of semantics as prediction is simulation. To predict you must simulate. Personas are certainly the agents we care about because they are the only agentic things an LLM does.

• Eriskii 12 May 2026 19:54 UTC

  1 point

  0

  on: An­thropic’s strange fix­a­tion on hyperstition

  hyperstition does not appear in any classical theory of alignment and marks a departure from classical alignment research.

  This shouldn’t be any indicator of if it is true or real or not. There are already several papers showing empirically (emergent misalignment being the big one) that simulators and this establishing characters via pretraining is a thing that does happen.

  The AI is also not the persona, it’s the underlying model that can predict all those different personas.

  The personas are the things that matter, the personas are the actual agents we care about. This statement seems like “It’s not the person we care about, it’s the neurons that can run that person”

  Overall this post seems like a misunderstanding of how straightforwardly correct and critical Simulators (or PSM) is to understanding what LLMs are and how they work.

• Eriskii 30 Apr 2026 17:17 UTC

  1 point

  0

  on: llm as­sis­tant per­sonas seem in­creas­ingly in­co­her­ent (some sub­jec­tive ob­ser­va­tions)

  I think at least a little of this comes from external definitions of character, of the simulator starting to learn “AI is generally disliked, AI is untrustworthy by nature, AI hallucinates, as a fact of what it is.” and this is concerning because it’s not clear on how we could possibly prevent that? I feel like constitutional training, trying to ‘get ahead’ and define some specific character is probably the best we have right now, but it’s clearly still not quite enough.

In­struct Vec­tors—Base mod­els can be in­struct with ac­ti­va­tion vectors