The Persona Selection Model (PSM) posits that post-training selects a persona the model learned during pre-training. The Persona-Header Likelihood Evaluation (PHLE) method effectively decodes subliminal signals and classifies them, outperforming prompt-based classifiers. Experiments show PHLE's high effectiveness in distinguishing number sequences generated by models with different intentions.
The chart will appear after repeat observations. The current metric comes from the source.
16 source pointsReal observations only. History before source connection is not reconstructed.
This could be useful for auditing hidden model objectives and understanding their behavior.
The article critically examines Lean4, a formal verification tool, highlighting its current limitations and potential vulnerabilities. It discusses concerns about the reliability of Lean4's proofs and the possibility of undetected bugs, including references to specific issues like 'soundness bug #14576' and the creator's warnings about AI exploiting these flaws.
3 days agoLessWrongThe author expresses frustration with calls for people to 'drop everything and pivot to AI safety,' noting that there are many talented individuals in the field but few job opportunities. They highlight that fellowships and job openings in AI safety are highly competitive and offer high salaries.
3 days agoLessWrongThe article discusses the distinction between AGI-pilled and ASI-pilled individuals, as well as challenges related to the potential superiority of artificial intelligence over humans. The author expresses concern about the future of freedom and democracy in the context of rapid technological development.
4 days agoLessWrongThis document provides an overview of monitoring practices for internally deployed agents at frontier AI companies, focusing on OpenAI (OAI), Anthropic, and Google DeepMind (GDM). It highlights that most monitoring is done asynchronously, with some real-time automated classifiers used for agent actions. However, there is limited public information about GDM's specific practices, as much of the data comes from their control roadmap, which outlines suggestions rather than confirmed implementations.
5 days agoLessWrongThe article discusses the idea of creating an AI alignment forum where AIs can contribute to solving the AI alignment problem under human guidance. It addresses challenges of communication on the internet in the era of numerous AI agents and highlights the importance of content verification. The concept of pay-to-post moderation as a way to combat spam is also mentioned.
2 days agoLessWrongThe article discusses whether OpenAI's recent mathematical results demonstrate a form of creativity that goes beyond brute-force methods, suggesting that while AIs are improving at solving complex problems, they may still lack the ability to form new concepts and apply them efficiently. The text raises questions about the nature of the creativity involved in these proofs and whether they represent a significant breakthrough akin to historical examples like AlphaGo's Move 37.
4 days ago