Or the thing learned the trick from some other book and tend to apply it.
The demonstration is flawed.
Not required to argue my argument. I tested the opposite and the models fail. Therefore, when the solution is not somewhere in enough numbers to stick while being trained, the AI doesn't work and walk the walk if deducting it.
