I have tried the method you suggested CaptainMaim and it seemed to have little results. The AAI Arm continued to prefer hovercraft over lvl 2 kbot/vehicle factories and it rates hovercraft factories over the others.
I believe training AAI only teaches it which units are the strongest at one factory (better unit choices), but the problem here lies in the fact that AAI doesn't always make the best factory choices (at least with the AAI I trained), which actually prevents it from making those better lvl 2 units in the first place. Choosing hovercraft factory over lvl 2 kbot/vehicle factories will cause it to make weaker hovercrafts on land maps (even if it was taught to make the best hovercraft unit choice) instead of inherently stronger lvl 2 kbot/vehicles.
From my experience, once AAI has decided to make a hovercraft platform, there is little chance that it would make a lvl 2 kbot/vehicle factory any time soon, since it seems to make a new factory only when it has developed to a certain point. If I can make a suggestion, Submarine, I think AAI should always prefer to make lvl 2 factories over lvl 1 ones when it has the capability/resources of supporting those higher tech structures, unless it is just starting out or it has low resources or all its factories have been destroyed (can't make lvl 2).
Another problem that I have discovered is that on some smaller maps, like Karst, AAI will constantly send newly created units to a certain place to "gather its forces" (on Karst this seems to be the middle left part of the map), but sometimes it continues to do that even if the enemy has already a stronger force in that same position. This results in AAI sending one unit at a time to the position already guarded by the enemy army and suiciding many units. For this reason I have observed that AAI loses faster to KAI on smaller maps, since it suicided its whole army and then get rushed. On bigger maps, AAI doesn't have this problem.
I think there are some things that AAI can learn from KAI and improve its early-mid game. For one, AAI seems to expand slower than KAI. While I am aware that KAI has vision of the whole map and therefore an advantage, I believe the way that AAI plays can be improved a bit. For instance, when KAI has already expanded to nearly the entire map, AAI has only expanded to around 1/4 of the map. When KAI expands it also builds new factories, often very close to the enemy's bases, and cranks out units from those factories. This is very effective and is one of the reasons why KAI rushes are so hard to stop. It would also be better if AAI prefers to crank out cheap units less and goes for strong units more (mostly lvl 1 Arm Kbots, I rarely see AAI make warriors). It would probably be better if AAI also techs faster to better buildings to give it more survivability, since it does not rush as aggressively as KAI does. Of course this is only my own opinions and there might be some way to adjust the AAI macro behavior that I am not aware of. If that is the case, forgive me and ignore what I just said.
Sorry this has become so long, originally it was only about AAI choosing hovercraft over lvl 2 factories. I'm trying to train AAI to improve its game so that it can perform better and maybe even win against KAI (for fun only). So far the record is 1 win 7 loss. AAI usually get rushed and dies before it reaches level 2. The only win AAI had over KAI was the one time when it sacrificed all tech as Arm and went 100% level 1 aircraft on WideOpenCombat and constantly bombed KAI Core's massed low hp wind generators (KAI's early reliance on those seems to be a weakness, but not a big one after it has advanced to higher tech energy sources).
For some reason, when playing on Mars in the top left starting position, AAI commander stops moving after making two solar collectors. Does this happen to anyone else? What is the cause of it?
Through several games I had some success making it build strong units, but there is one thing I had trouble teaching it. 
