> They will not foresee/anticipate events, that are unlikely or non-existent in their training data, but are bound to happen due to real world circumstances. They are not intelligent in that way.

Can you be a bit more specific at all bounds? Maybe via an example?