For computational optimization, the von Neumann game theory terms are applied to actual logic systems and we end up with computer vision optimizations and video/audio codec best guess optimizations, or some interesting applications of evolutionary computing [4] in verifiable applications like signal optimization in antenna design [5] or industrial manufacturing techniques and structural integrity or architectural optimizations via generative design [6] in software [7] like generative computer automated design, ie Fusion 360 or Grasshopper [8].
Rudimentary modeling with von Neumann game theory fails to account for variables in reality, like the ambiguity effect and in applications like optimization of social choices like economic modeling or voting i.e. Gibbard's theorem. There's often situations where game theory fails to account for an adversarial evasion of the rule set - https://en.wikipedia.org/wiki/Metagame_analysis
If the primitive von Neumann game theory models were used to train an artificial intelligence system that controlled military weapons, there would have been global thermal nuclear winter in several decision making "games" like the 1983 Soviet nuclear false alarm situation.
https://en.wikipedia.org/wiki/1983_Soviet_nuclear_false_alar...
[1] https://en.wikipedia.org/wiki/Ambiguity_effect
https://en.wikipedia.org/wiki/Ambiguity_aversion
[2] https://en.wikipedia.org/wiki/Behavioral_game_theory
[3] https://en.wikipedia.org/wiki/Bounded_rationality
[4] https://en.wikipedia.org/wiki/Evolutionary_computing
[5] https://en.wikipedia.org/wiki/Evolved_antenna
[6] https://en.wikipedia.org/wiki/Generative_Design
[7] https://fab.cba.mit.edu/classes/865.18/design/generative/ind...
[8] https://www.engineering.com/DesignSoftware/DesignSoftwareArt...