I think you have a wildly unrealistic idea of what AI can do.

Modern AI is mostly machine learning. This allows you to do things like recommendation and classification systems; it’s amazing for getting statistical patterns out of a large collection of uniform data; it is in no way a system that performs reasoning.

For example, how would you conceivably use AI to deal with “geopolitical cooperation and conflict”?

First, you’d need a large corpus of examples geopolitical problems with solutions and some sort of score for that solution. How would you even represent a geopolitical problem in any sort of way that was useful to a computer program? How would you agree what the score was? How many examples could you come up with? You need thousands at least to train a machine learning model.

Even if you somehow accomplished this — which seems impossible — what use would it be? At the end, it would simply spit out a result. There would be no explanation of why it had come to that result, because machine learning never answers the question “why”. Who would be convinced by a simple statement — “Do this” — without any possible explanation as to why you should do that?

Modern AI in the form of machine learning is a powerful statistical tool that gets results undreamt of twenty years ago, but it’s not some magic bullet that can solve the world’s problems. And in the last fifty years, we’ve made almost no progress on general AI, systems that can reason — which is what you really want for real-world problems like this.