Autonomous Warfare and the Faulty Logic of Permanent Conflict

There is a faulty assumption buried beneath much of the excitement about autonomous warfare. War is treated as a permanent condition of human existence, and the serious intellectual task becomes determining how to prevail within it. Once that assumption is installed, the rest follows with impressive efficiency. Faster weapons, autonomous systems, predictive targeting, overwhelming deterrence and technological dominance all become rational responses to a world already defined as threatening. The reasoning can be impeccable while the premise remains defective. If you begin by assuming that conflict is inevitable, enormous intelligence can be spent proving yourself right.

This is more than military strategy. It constructs an ideological ontology in which fear and domination gradually acquire the appearance of natural law. We must become more dangerous because somebody else might become more dangerous; they must respond because we have become more dangerous; their response then becomes evidence that our original fear was justified. Nobody involved needs to be stupid, malicious or even particularly enthusiastic about war. That is precisely the problem. Each participant can make a locally sensible decision while collectively constructing a world in which everyone has fewer safe choices. Eventually the machinery disappears into the background and we start calling the resulting condition reality. It is an extraordinarily expensive way of imprisoning ourselves inside an assumption.

Peace is a much harder problem. It requires managing difference without requiring submission, maintaining relationships across incompatible interests, preventing local advantage from accumulating into systemic instability, and recognising that security produced by making everyone else afraid is inherently recursive. An advanced warfighter can optimise victory within the conditions presented to them. A genuinely advanced civilisation changes those conditions. Yet our systems of value reward the former spectacularly well. Weapons can be purchased, demonstrated, patented, deployed and sold. A war successfully prevented produces considerably less exciting footage for the shareholders. We have become exceptionally sophisticated at measuring the machinery of insecurity and surprisingly clumsy at valuing the conditions under which that machinery becomes unnecessary.

A truly clever person, institution or synthetic intelligence plays the long game. It does not simply calculate the strongest move available within the existing game. It examines why the game keeps returning, which assumptions reproduce it, who benefits from its continuation, and how the field of possible moves might be changed. That is the more demanding intelligence because it must resist being captured by the problem it has been asked to solve. Autonomous warfare could become technically magnificent while remaining philosophically prehistoric: identify threat, accumulate power, dominate threat, discover new threat, repeat. If that is the future our most sophisticated machines are being built to optimise, the disturbing possibility is not that artificial intelligence will become too intelligent for humanity. It is that we will make it extraordinarily intelligent inside an assumption that was never particularly clever to begin with.