just ask yourself: what would a "thinking war machine" actually "think" about? it's not as if war is just a boardgame - heck, it's not as if the political and military moves we make are even carefully thought-out at all!
In fact, war itself is well-known to be fundamentally irrational. There's even something in economics called the "war puzzle" or "war problem": under the economic model of rationality, war is irrational.
Actors can always generate better outcomes by negotiation, and in real-world case studies typically both sides believe they have a much greater than 50% chance of winning (which violates the law of conservation of probability...)
As Clausewitz might have said if he'd known about Darwin: war is reproductive competition carried out by other means.
As such, creating bigger and bigger machines to prosecute wars is the stupidest thing humans could possibly do. On the other hand, if you think a weapon is a tool for changing your enemy's mind, then machines that educate are the most powerful weapons of all.
If we want to dump billions into making the world safe for American Imperialism, teaching machines of the kind envisioned in "The Diamond Age" would be a far better investment than exa-scale hardware that won't be able to think, but will be able to knock one more decimal place of uncertainty off of opacity coefficients for thermonuclear simulations.
But human beings are too stupid and irrational to do that, and would far prefer to engage in the least efficient, least effective strategy for solving any human problem: war.
There are people who are so stupid that they believe, for example, that because war was required to end slavery in the US that it was somehow a good solution, and they are so ignorant that they are unaware that slavery was eliminated in many other places without warfare. Simply because some bunch of idiots somewhere were too stupid to solve their problems without war doesn't mean that war should be the go-to solution for any problem that faces us.