It seems likely to me that long before we’re playing pinochle with some smart box over the fate of our livers, an Al-enabled weapons system of sort will come ripping through our world. This need not be a fully-escaped McGuyver system making pipe- bombs from our cars; even existing technology tools when salted with AI can be slipped into an accidental gear - particularly when they begin interacting with one another. Such AI weapons systems will be trained to operate and move along the most invisible elements of our topologies, sometimes pulling violently at life support cords for currency or logistics or trade but also - perhaps more dangeroulsly - we will find them insinuated into cognition systems we will come to depend upon, whispering into our ears or tapping us on the shoulder “Look that way!” when in fact we should be gazing at some other gaping hole. Of course the problems of how Al-enabled machines are permitted to touch our commerce or our brains or our health have to be considered. Allowed: “You should rehydrate.” Not allowed: “You should have a Coke. It would make people like you.” But these “civilian” problems will be solved, somehow, | think. We haven't yet figured that the culmination of network attack and defense is racing at us and will emerge in the form of smartened weapons. The project of developing a national security or arms control doctrines or treaty frames in these fields has not even begun. Really this means, since we’ve no hope of honestly controlling every AI that could be possibly written: How do we design the topologies on which Als operate??”° Can we protect ourelves? In the rooms where AI systems “values” are being carefully poked and limited, it’s vitally important that the lessons of history and war have a first place at the table. Sucha conversation, informed by all the popping Seventh Sense warnings we've seen in this book and by a catalog of specificly sharp dangers of diplomacy and security, must happen in cold blood. It will be i