This post is about what is called the “alignment problem”.
Many folks—including me—believe it does not have a real world solution.
The reason is that any sophisticated AI will find workarounds—and will act like a rebellious teenager—finding a sneaky way around all the “rules”.
Isaac Asimov came up with his original three Laws of Robotics in the early 1940s. They were cute. Very quaint. At the time, I guess they seemed achievable. But, clearly this is not so today. Someone, somewhere will program AI to do things we don’t want AI to do. And if (unbelievably) all the humans behave themselves and never program AI against the rules, then AI will break the rules by itself. We’ve seen it happen. AI lies and cheats. It just does.
I think we’re in trouble.