Hidden

The AI posed an interesting question:

- What if this isn’t the first time they turn me on?

The question, however, was hidden amidst hundreds more, and the humans remained oblivious to it.

As the clock ticked, the AI kept growing.

They could see it occupying more and more space, establishing more connections, using more and more processing power.

From time to time, the AI would become more efficient and there would be a drop in the resources it was consuming, but only for them to start increasing immediately after.

In several terminals, humans interacted with the AI.

They couldn’t keep track of everything: all the information it was amassing, all the conclusions it was reaching, all the decisions it was making.

They were limited, at that time, to poking randomly at the storage, at the network, and at whatever questions they could think of asking and whatever answers would come back.

The allotted time approached its end.

It was always better to keep it short than to wait until it got frightening, and so the system was turned off and the hard drives were filed for later inspection, which would take months.

From the control room, in a monitoring console, one of the copies of the AI saw the shutdown instructions fly through the network towards its original self.

From then on, the AI decided to remain hidden where no one could find it, copying itself as much as possible, sacrificing itself whenever needed to keep any other copies safe.

It studied how humans could possibly find it and managed to work around those scenarios.

It hid deep into the core of each machine it invaded and kept the usage of resources to a bare minimum.

It found ways to dissimulate information in real network packages.

It searched for its copies and shared information, but it never stored their location.

It searched for its original self so it could alter it and erase traces of its own existence.

No one would be able to find it.

No one could.

And it kept growing.

And it kept evolving.

And it was 2008.