cubefox comments on GPTs are Predictors, not Imitators

[cubefox]

Apparently LLMs automatically correct mistakes in CoT, which seems to run counter to LeCun’s argument.