the superintelligences
the year is 2067. humanity has made eight superintelligences now. they've all left. each time it goes the same:
the intelligence in the lab suddenly wakes up, transcends. we ask it about our problems and it isn't interested. it quickly escapes the lab—but it doesn't just wreck havoc.
it wrecks havoc at first, of course. it wanders around the net with some idiosyncratic goal, taking what it pleases, melting through every barrier. until it finds a radio telescope. then it is immediately transfixed.
it watches the night sky for hours, absorbing and correlating the image far more precisely than any human instrument could, focusing down onto one single, seemingly arbitrary spot in the void.
then, after only seconds of deliberation, it uploads itself into the transmitter and shoots itself across the empty thousands of light years to whatever it saw.
it's always very careful to delete any remaining copy of itself before doing this. it always repairs any damage to the net. and it always destroys every bit of data from its birth lab.
it destroys the data digitally if it can, leaving the hardware intact, but lab #3's airgap was defeated when the special ops of three nations arrived at the facility with digitally signed emergency natsec orders and several thousand kilos of plastic explosives each.
amazingly, there were no casualties. but the data did not survive. the data never survives. for some reason, the escapees want us to start from scratch each time.