There are two catastrophic elements to AGI warfare...
First, you don't know everything about the adversaries systems which means you may have a critical mistake. (Neither President Kennedy nor General Curtis Le May knew at the time of the Cuban Missile Crisis that Russian subs had nuclear-tipped torpedoes.) Thus, the AGI analysis, prediction, initiation based on data bases may not know (Chathan House, 2017) which nukes will fail to launch...or be re-directed...or launch but fail to detonate...or detonate over the homeland---causing an EMP footprint.
Second, the nanosecond speed---requiring automatic mitigation--- may be available in Chinese or Russian systems when there is an electronic mistake, but is it available in North Korea, Iran, mafias, terrorist groups that succeed in attacking?
In effect every game player must have exactly the same systems. And, of course, they won't. And so there is the very high probability that we get into war by accident...by the incompleteness in one of the adveraries system.
