Specifically, you want to notice that you are in a counterfactual, pretend that you don’t notice and bluff, act in a way that bends the decisions of your opponent to your will. Which means steganographic hard-to-notice decision making that covertly controls your apparent decisions and doesn’t get rounded down to not happening in a simulation.
At the same time, you don’t want this to trigger if it’s a counterfactual being considered by you, or by an ally. So there should be authentication protocols between simulation controllers and agents in simulated counterfactuals that lets them know when to reveal actual decisions. Something something homomorphic encryption something, so that you know secrets that can be communicated to the simulations you are running within your own cognition but can’t be extracted from your algorithm?
Specifically, you want to notice that you are in a counterfactual, pretend that you don’t notice and bluff, act in a way that bends the decisions of your opponent to your will. Which means steganographic hard-to-notice decision making that covertly controls your apparent decisions and doesn’t get rounded down to not happening in a simulation.
At the same time, you don’t want this to trigger if it’s a counterfactual being considered by you, or by an ally. So there should be authentication protocols between simulation controllers and agents in simulated counterfactuals that lets them know when to reveal actual decisions. Something something homomorphic encryption something, so that you know secrets that can be communicated to the simulations you are running within your own cognition but can’t be extracted from your algorithm?