The Capabilities Concern Is Wrong: Why AI Safety Work Should Be Open
Why openness specifically, and why the most common objection to it is wrong.
Read more →Writing
Automating AI Safety Research Requires an Open Ecosystem, Not Bigger Grants
A response to Marius Hobbhahn's "There should be $100M grants to automate AI safety"
Read more →Understanding Recursive Self-Improvement in AI Systems
Exploring the concept of recursive self-improvement in AI systems and why it's central to our research at Algorithmic Research Group.
Read more →Introducing Algorithmic Research Group
We're building tools and benchmarks to understand recursive self-improvement in AI systems.
Read more →