Sorry, I’m still a little confused. If we establish an AI’s terminal goal from the get-go, why wouldn’t we have total control over it?
We don’t know how to do that. It’s something that falls out of its training, but we currently don’t know how to even predict what goal any particular training setup will result in, let alone aim for a specific one.
Sorry, I’m still a little confused. If we establish an AI’s terminal goal from the get-go, why wouldn’t we have total control over it?
We don’t know how to do that. It’s something that falls out of its training, but we currently don’t know how to even predict what goal any particular training setup will result in, let alone aim for a specific one.