AI SAFETY & ETHICS

Sanity-checking “Incompressible Knowledge Probes”

LessWrong AI

Or, did a chief scientist of an AI assistant startup conclusively show that GPT-5.5 has 9.7T parameters?