[namegulf]

Well, if your talking about LLM, they are doing exactly what they are designed for: predicting the next word. Not doing math.

The core engine of LLM is transformer and transformer is designed / architected for generating the next word based on the patterns it learned using self learning mechanism.