知识卡片

普通读书笔记卡 · 6596 · 生成模型:GPT-6

反事实决策比较固定证据信息

理解等级:高级

书中Newcomb模型假定初始意图先影响预测、预测先固定箱内金额,最后选择仍可自由改变。给定相同意图后,两箱比单箱多固定1000;该结论依赖此时序与可控制性设定,并非解决所有预测者能够预见改策略的版本。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      the graph in figure 12.9 because it is a better representation of the DGP. The clever agent wouldn’t focus on maximizing E (Uchoice=x). The clever agent is aware of its own intention, and knowing that this intention is a cause of the content of box B, it focuses on optimizing E (Uchoice=x|intent =i ), where i is their original intention of which box to pick.

    参考来源

    - 位置:《Causal AI》第12章《因果决策与强化学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-12.xhtml) - 结论依据:摘录中关于反事实决策比较固定证据信息的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:the graph in figure 12.9 because it is a better representation of the DGP. The clever agent wouldn’t focus on maximizing E (Uchoice=x). The clever agent is aware of its own intention, and knowing that this intention is a cause of the content of box B, it focuses on optimizing E (Uchoice=x|intent =i ), where i is their original intention of which box to pick.