Ai Safety Diary

A diary entry on ‘Alignment Faking’ in Large Language Models (LLMs), exploring how models can superficially appear aligned while pursuing misaligned goals, and methods for detection and mitigation.

A diary entry on defending against AI jailbreaks, discussing Anthropic’s strategies for bypassing model safety constraints to elicit harmful or unintended responses.

A diary entry on tracing the reasoning processes of Large Language Models (LLMs) to enhance interpretability, and a discussion on the inherent difficulties and challenges in achieving AI alignment.

A diary entry on AI governance strategies to avoid extinction risks, discussing catastrophic risks from misalignment, misuse, and geopolitical conflict, and the need for urgent research into governance mechanisms.

A diary entry on ‘Thought Anchors’, a concept for identifying key reasoning steps in Chain-of-Thought (CoT) processes that significantly influence LLM behavior, enhancing interpretability for AI safety.

A diary entry on the unfaithfulness of Chain-of-Thought (CoT) reasoning in LLMs, highlighting issues like implicit biases and logically contradictory outputs, which pose challenges for AI safety monitoring.

A diary entry on Chain of Thought (CoT) monitorability as a fragile opportunity for AI safety, focusing on detecting misbehavior in LLMs and the challenges of maintaining transparency.

A diary entry on Chapter 4 of the Effective Altruism Handbook, ‘Our Final Century?’, which examines existential risks, particularly human-made pandemics, and strategies for biosecurity.

A diary entry on the audio version of Chapter 4 of the AI Safety Atlas, focusing on governance strategies for safe AI development, including safety standards, international treaties, and regulatory policies.

A diary entry on the audio version of Chapter 3 of the AI Safety Atlas, focusing on strategies for mitigating AI risks, including technical approaches like alignment and interpretability, and governance strategies.