AI safety and consciousness — page 1
- AI 2027 (Kokotajlo)
url:ai-2027.com· 2025-04-03 · metadata only - An Extremely Opinionated Annotated List of My Favourite Mechanistic Interpretability Papers v2 (Alignment Forum)
url:alignmentforum.org/posts/nffst5mio7bcaqhpa/an-extremely-opinionated-annotated-list-of-my-favourite-1· 2024-07-07 · metadata only - Attribution graphs, the biology of the model
url:transformer-circuits.pub/2025/attribution-graphs/biology.html· 2025-03-27 · metadata only - Circuit tracing / attribution graphs (methods)
url:transformer-circuits.pub/2025/attribution-graphs/methods.html· 2025-03-27 · metadata only - Consciousness in Artificial Intelligence: Insights from the Science of Consciousness
arxiv:2308.08708· 2023-08-17 · text available - Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teaming
arxiv:2501.18837· 2025-01-31 · text available - Could a Large Language Model be Conscious?
arxiv:2303.07103· 2023-03-04 · text available - Detecting Strategic Deception Using Linear Probes
arxiv:2502.03407· 2025-02-05 · text available - Emergent Introspective Awareness in Large Language Models (Anthropic, October 2025)
url:transformer-circuits.pub/2025/introspection· 2025-10-29 · metadata only - Is Evaluation Awareness Just Format Sensitivity? Limitations of Probe-Based Evidence under Controlled Prompt Structure
arxiv:2603.19426· 2026-03-19 · text available - Is Power-Seeking AI an Existential Risk?
arxiv:2206.13353· 2022-06-16 · text available - Mapping the mind of a large language model (Anthropic)
url:anthropic.com/news/mapping-mind-language-model· 2024-05-21 · metadata only - Optimal Policies Tend to Seek Power
arxiv:1912.01683· 2019-12-03 · text available - Palatable Conceptions of Disembodied Being
arxiv:2503.16348· 2025-03-20 · text available - Parametrically Retargetable Decision-Makers Tend To Seek Power
arxiv:2206.13477· 2022-06-27 · text available - Power-seeking can be probable and predictive for trained agents
arxiv:2304.06528· 2023-04-13 · text available - Reasoning Models Don't Always Say What They Think
url:anthropic.com/research/reasoning-models-dont-say-think· 2025-05-08 · metadata only - Situational Awareness (Aschenbrenner)
url:situational-awareness.ai· 2024-05-29 · metadata only - Sparse Autoencoders Find Highly Interpretable Features in Language Models
arxiv:2309.08600· 2023-09-15 · text available - The Value Axis: Language Models Encode Whether They're on the Right Track
arxiv:2606.17056· 2026-06-15 · text available - Tracing thoughts of a language model / Circuit tracing (Anthropic)
url:anthropic.com/research/tracing-thoughts-language-model· 2025-03-27 · metadata only - Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision
url:openai.com/research/weak-to-strong-generalization· 2023-12-14 · metadata only - When Activation Oracles Learn Not to Read: Concept-Specific Blind Spots in Fine-Tuned Oracles
arxiv:2607.23379· 2026-07-25 · text available