I don’t see how this amounts to central control. At best it is parallel predetermination, but that breaks down because the actions of the AI are determined by the environment, not the utility function alone. Central control implies two-way communication and is impractical when the latency is measured in decades.
I don’t see how this amounts to central control. At best it is parallel predetermination, but that breaks down because the actions of the AI are determined by the environment, not the utility function alone. Central control implies two-way communication and is impractical when the latency is measured in decades.