This article examines emergent misalignment in the recurrent model Ouro-1.4b using a logit lens. The author demonstrates that misalignment is inducible and its representation varies across loops. Experiments were conducted using LoRA and evaluated with Sonnet 5.5 subagents.
The chart will appear after repeat observations. The current metric comes from the source.
8 source pointsReal observations only. History before source connection is not reconstructed.
Studying emergent misalignment is important for understanding the safety of recurrent LLMs.
The text discusses the debate between 'Alignment Engineering' and 'Misalignment Science' in AI alignment research, highlighting concerns that current alignment methods might inadvertently accelerate capabilities development, increasing risks of rapid technological growth. It questions the opportunity costs of focusing on alignment engineering and proposes a shift towards 'Misalignment Science' as a more robust approach.
5 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.
2 days agoLessWrongThe 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.
yesterdayLessWrongThe 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.
yesterdayLessWrongThis 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.
3 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.
2 days ago