The question is inability of ChatGPT to do logical reasoning. Multiplying numbers is a simple test to prove this.

The same with coding: it can produce code for tasks where something similar was in training dataset, but if you ask to do completely unknown problem - it will fail.