Sure it is completely impractical for humans, but it is theoretically possible, and might even be practically possible for an AI running on a quantum computer." Alright, but tempering with the memory of the agent is effectivelydestroying the agent: it invalidates the fundamental assumptions of it...