Let's get back on topic.
@Schulz AI development is one of the major features of TripleA and is a priority. Unfortunately it does take a ton of time and labor to do most anything in the code base, it's old and a bit hacky.
I would recommend if there are specific situations where the AI could make a better decision, posting a game save and describing what the AI did and could have done for a better decision could be very useful. Specific spot examples like that we can probably look at specific parts of the code and perhaps improve them..
In theory as well the code architecture tries to support "plug-and-play" AI modules. Namely you can create your own AI and drop it in. It's not done as cleanly as it could be, requires a good bit of programming knowledge, and kinda is a big effort code-wise. Perhaps it's best for now to collaborate to improve the "pro AI".
@alkexr RE: machine learning, the AI could play itself and generate a dataset that way. Unfortunately short term that is not very feasible as the code is in a messy state which makes big changes or additions prohibitive.