x-risk-themed
Sometimes, a friend who works around here, at an x-risk-themed organisation, will think about leaving…
Category: Opinion
Toward a Better Evaluations Ecosystem
Model evaluations are broken. Numbers that are often cited alongside one another as evidence of…
Model Spec Midtraining: Improving How Alignment Training Generalizes
tl;dr We introduce model spec midtraining (MSM): after pre-training but before alignment fine-tuning, we train…
Positive Feedback Only
This story was written collaboratively with Claude. I brainstormed ideas with it and decided what…
What if LLMs are mostly crystallized intelligence?
SummaryLLMs are better at developing crystallized intelligence than fluid intelligence. That is: LLM training is good…
Decision theory doesn’t prove that useful strong AIs will doom us all
Bottom-line up frontTraining for optimal behavior doesn't inevitably lead to act-utilitarian world optimizers ("WorldSUM agents").…
Psychopathy: The Mechanics
Note on LLM useThis sequence is based on hundreds of hours of literature research and…
The AI Ad-Hoc Prior Restraint Era Begins
The White House has ordered Anthropic not to expand access to Mythos, and is at…
Your rights when flying to Europe
Europe (and the UK) have strong protections for flyers in the case of delayed or…
![[Linkpost] Interpreting Language Model Parameters](https://res.cloudinary.com/lesswrong-2-0/image/upload/v1777975579/lexical_client_uploads/tmiqca6oiswnadtrjz52.png)
[Linkpost] Interpreting Language Model Parameters
This is the latest work in our Parameter Decomposition agenda. We introduce a new parameter…