知识卡片

普通读书笔记卡 · 8690.d · 生成模型:GPT-6

可靠性目标取决于失败后果

理解等级:中级

总体99%成功不能描述失败是否集中于关键操作,也不能证明多步骤整体可靠。需按故障类型和后果定义可接受范围,并区分可恢复错误、不可逆操作及需人判断的情况;平均准确率不足以决定自主程度。

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

      99% of the time is still not good enough, especially when the failures are unpredictable and the 1% of failures can be potentially catastrophic. The 99% problem is also the one that has long plagued self-driving cars and prevented their broader adoption. This doesn’t mean we can’t deploy autonomous LLM agents; we just need clever product design that can shield the user from their failures.

    参考来源

    - 位置:《Designing Large Language Model Applications》第10章《外部工具与智能体》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch10.html) - 结论依据:摘录中关于可靠性目标取决于失败后果的定义、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:99% of the time is still not good enough, especially when the failures are unpredictable and the 1% of failures can be potentially catastrophic. The 99% problem is also the one that has long plagued self-driving cars and prevented their broader adoption. This doesn’t mean we can’t deploy autonomous LLM agents; we just need clever product design that can shield the user from their failures.