Deconfusing Corrigibility

Claude AI discovered that NIST's ECDSA cryptographic seeds were generated from Unix timestamps using lrand48 RNG, solving a decades-old mystery in cryptography.

Analyzes how undisclosed internal AI models create dangerous information gaps between developers and the public, potentially misleading regulators about true frontier capabilities.

Projection sampling preserves alignment gains from fine-tuning while reducing the capability loss typically associated with safety-focused training.

Mercor Research found that 90% of US working hours involve tasks with little to no coverage in existing AI benchmarks, making it difficult to assess model performance on real-world work.

Explores connections between imperfect recall games, anthropics, decision theory, and reinforcement learning through the Sleeping Beauty problem framework.