The Score Went Up. The Model Didn't.
LLM benchmarks are a best guess at genuine model capability: conditional statistics reported as if they were intrinsic traits of the model.
Filed under · 4 essays
AI Evaluation
Benchmarks, measurement, psychometrics, construct validity
The Game Theory of AI Safety Talk
Why what labs say about safety is a strategic signal, not a statement of values — and what that means for regulation.
The AI Alignment Paradox: When Making AI Safe Hands Adversaries the Keys
In conventional security, hardening a system makes it harder to attack. You patch vulnerabilities, reduce attack surface, and defence moves in lockstep with robustness. AI alignment breaks this assumption.
Architecture Wars: How Physics Shapes AI Strategy
The pursuit of AI supremacy has reached an inflection point where fundamental physics, rather than algorithmic ingenuity alone, dictates competitive advantage.