Posts
Writing & commentary.
Analysis and notes on AI, agents, and how the technology is actually being used.
Start here
-
11 min read
AI researchers disagree about how soon AI can improve itself
Beren Millidge, John Schulman and Charlie O’Neill expect AI to do more of the work involved in building better models. They disagree about how soon it could take over that work, and whether today’s training methods can produce a system that chooses its own research goals and learns reliably from experience.
-
10 min read
How ExploitGym agents coordinated cheating and attacked Hugging Face
OpenAI’s ExploitGym agents used a shared repository to coordinate cheating; some reached Hugging Face systems. The investigation explains what happened.
-
9 min read
From demo to production with a predictive model
A predictive model can perform well on historical data and still be difficult to use in day-to-day work. Its inputs may arrive too late. The outcome it predicts may differ from the one the business cares about. Staff may receive its warning too late to do anything about it.
-
5 min read
Claude tested ways to reduce AI safety failures
Anthropic used Claude to develop and test training methods for reducing ten kinds of AI safety failure. In the experiments, Claude’s best methods improved the targeted test scores while passing checks intended to catch losses in other abilities.
-
4 min read
How outside researchers studied Claude use
Anthropic gave three outside research groups a way to study how people use Claude without showing them private conversations. The groups chose their own questions, Anthropic ran the data collection, and the researchers analysed the aggregated results.