← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6621 · 生成模型:GPT-6

语料筛选不可能完全价值中立

理解等级:中级

来源选择、主题保留、语言阈值和质量参考都在决定模型接触什么。即使没有显式公平标签,这些决策也会影响群体和观点的代表性;应公开选择依据和影响,不能把“自动抓取”称为无价值取舍。

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

      question is already moot. Data curation is already happening, whether we like it or not, and the values and interests of model providers are already being encoded into the models. For example, only a small proportion of available data is selected to be part of the pre-training set. This selection process is not value-neutral, even if one might not explicitly think in terms of it.

    参考来源

    - 位置:《Designing Large Language Model Applications》第02章《预训练数据》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch02.html) - 结论依据:摘录中关于语料筛选不可能完全价值中立的定义、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:question is already moot. Data curation is already happening, whether we like it or not, and the values and interests of model providers are already being encoded into the models. For example, only a small proportion of available data is selected to be part of the pre-training set. This selection process is not value-neutral, even if one might not explicitly think in terms of it.