We need Large Language Models (LLM) for Air warfare
We know that LLM's are like really sophisticated autocomplete. Basically all text files(books, internet) are included in a language model. if you say what is on a race track? It could predict that most things written about that subject replied with a "car".
With a Math based visual LLM variation, I think we could predict from previous dogfights of what path an Aircraft took to know what path we can take. E.g: Radar detects an object at certain x-y-z axis with nose pointing to north then we can predict where we can go or do based on previously won dogfights.