One of the many things that annoys me about modern AI technology is if it did somehow destroy the world as its early boosters often (and bizarrely) claimed it might, it's far more likely to do so by accident than design.
There'd be no cannily launching a nuclear attack against Russia knowing that the counterattack would destroy its enemies at home. Instead, it would likely happen because someone asked for a history of nuclear tests and it got the wrong end of the stick. 'In retrospect, my decision to unleash our entire nuclear arsenal against China was ill-advised,' it would wheedle in that infuriating little-birthday-boy tone, as you closed your eyes to the encroaching mushroom cloud, waiting for the shockwave to hit.
