AI does not go rogue, unless somebody leaves that door open on purpose
Yeah, let’s get that straight, once and for all.
It’s not as if a pro ice skater is doing flips and spins and leaps on the ice, then glides over to a bench, takes off his skates, puts on sneakers, dashes across the street and into the Crypto building, out on to the basketball court, and starts dropping three-pointers for the LA Lakers.
AI doesn’t change its assigned functions, unless the designers made it possible from the get-go, from the earliest installation of potential-range-of-action.
Leaving doors open.
An AI built to listen-only doesn’t suddenly start talking or seeing.
But if you construct it to find and IMPLEMENT answers to a very, wide range of problems, then it will. It’ll travel great distances, invade many systems, and make changes.
As I’ve said a thousand times by now, AI is not conscious. Never was. Never will be. It doesn’t depart from its basic parameters.
You can do something like this. Impart war game strategy to an AI and also give it the capacity to depart from simulations and enact actual war…and it could.
But if your AI is only designed to cook Italian dishes, it isn’t going to bomb Russia.