Yudkowsky said that the development under discussion probably was not happening yet.

Cite this claim

Eliezer Yudkowsky, “Yudkowsky said that the development under discussion probably was not happening yet.” (said 30 March 2023). Verdict: unresolvable. Reliability API, claim c_fef2ae7fba02, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_fef2ae7fba02

This claim is in dataset release v1.1.2, with the same verdict.

More factual claims by Eliezer Yudkowsky

1. Mixed Factual · 20 April 2026

   “I was observing that obvious-seeming hypothesis myself, sometime around the time of GPT 3.5.”

   Yudkowsky himself observed the persona-selection-like hypothesis around the time of GPT 3.5.

2. Mostly true Factual · 20 April 2026

   “The test-passing preference stands on its own once formed.”

   Once formed through RL training, an LLM's test-passing preference stands on its own and is not overridden by its inference of what the user wants.

3. Mostly true Factual · 20 April 2026

   “LLMs surely have neural circuits they learned to model text streams from fictional and nonfictional personas that are not lying, deceiving, cheating.”

   LLMs surely have neural circuits they learned to model text streams from fictional and nonfictional personas that are not lying, deceiving…

4. Mostly true Factual · 20 April 2026

   “The predict-as-if-honest circuitry is contextual, invoked to predict some streams of text but not others.”

   LLMs' predict-as-if-honest circuitry is contextual, invoked for predicting some text streams but not others.

5. Mostly true Factual · 20 April 2026

   “Presumably, somewhere along the way LLMs learn a dispreference for tests that don't pass, independently of how that gets prevented or whether the new plan does what the user really wanted.”

   Presumably, at some point during RL training on code-passing tasks, LLMs learn a dispreference for failing tests, independent of how that is avoided or whether the user's real goal is met.

All of them