Complex plans must be penalized in the same way that complex theories must be (a plan is really just a theory that a certain course of action will lead to a good outcome).
Mm, no. Theories are mutually exclusive. The thing that makes a complex theory unlikely is the fact that it has 10^N, for large N, other theories to compete with. By your definition of a plan, plans are not mutually exclusive, so this analogy vanishes. Of course, you could define a plan as the theory that a certain course of action will lead to the best outcome out of all possible plans, in which case your statement would be true, but wouldn’t apply: the agent that finds the best possible plan before acting starves to death, decays, and is forgotten long before ever eating the food in front of it.
I find this comment confusing. Represent a plan as a list of discrete actions. Then “X,Y” is different from “X,Y,Z” where X/Y/Z are actions. Picking “X,Y,Z” means you did not pick “X,Y”. The number of plans is also exponential in the number of actions, so a more complex (longer) plan has an exponentially greater number of competitors.
You defined a plan as a theory that a certain course of action will lead to a good outcome. Picking “X,Y,Z” does not mean you don’t think “X,Y” will also lead to a good outcome. Therefore, they aren’t really competitors.
Mm, no. Theories are mutually exclusive. The thing that makes a complex theory unlikely is the fact that it has 10^N, for large N, other theories to compete with. By your definition of a plan, plans are not mutually exclusive, so this analogy vanishes. Of course, you could define a plan as the theory that a certain course of action will lead to the best outcome out of all possible plans, in which case your statement would be true, but wouldn’t apply: the agent that finds the best possible plan before acting starves to death, decays, and is forgotten long before ever eating the food in front of it.
I find this comment confusing. Represent a plan as a list of discrete actions. Then “X,Y” is different from “X,Y,Z” where X/Y/Z are actions. Picking “X,Y,Z” means you did not pick “X,Y”. The number of plans is also exponential in the number of actions, so a more complex (longer) plan has an exponentially greater number of competitors.
You defined a plan as a theory that a certain course of action will lead to a good outcome. Picking “X,Y,Z” does not mean you don’t think “X,Y” will also lead to a good outcome. Therefore, they aren’t really competitors.