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?