Unfortunately, the author's background and experience is highly relevant to the ideas presented in the article. Someone with no training or practical experience in air combat who is proposing "innovation" in that area is like virgins writing papers on how to improve their readers' sex techniques.
For example...his "Figure 1" image of a machine performing VID based on shape recognition in some EO sensor. Great idea...except any pilot who has that same view of a full-frame planform of your adversary in the HUD could make that same identification with probably the same level of accuracy. More importantly, someone who has actually engaged in air combat knows that such views are extremely rare and reflect something that is a component of maybe 1% of the BFM/ACM/ACT scenarios out there.
Again, it is like someone who thinks that every girl loves anal and a facial "money shot" because that's what seems to be in all the porn they've watched.
Conceptually, he's not wrong; eventually there will be technology that will be able to have the SA to make autonomous decisions faster and better than humans currently do. Machines will be better able to ID, better assess range/position/energy/maneuvers of an adversary, better know ownship energy state and capabilities, and will be able to more perfectly select the correct tactic, and execute the control inputs to more perfectly execute the maneuver, time the shot, etc. Everything will be better than what a standard 1-each human could do.
That time is just not now.