</head><body><nav><div class="control-group"><div class="control-item"><div class="control-item__button-row"><a class="control-item__button control-item__button--size-main" href="/?commit=26db3b2b20aec5cffcda11a6ff35f6e13f0153759350fb5bfe8cd68e3d600516"><span class="">freeze</span></a></div><div class="control-item__title-overlay control-item__title-overlay--interactive"><span class="control-item__data-copy--end">latest universe</span></div></div></div></nav><main><h3>Minimal object containing this commit</h3><div><a class="visit" href="6f255651b9b007a6849a55f0ba47de341f034862b6e5ebbe150830128adaf189?commit=26db3b2b20aec5cffcda11a6ff35f6e13f0153759350fb5bfe8cd68e3d600516">6f255651b9b007a6849a55f0ba47de341f034862b6e5ebbe150830128adaf189</a></div><h3>Commit Diff</h3><pre>commit 26db3b2b20aec5cffcda11a6ff35f6e13f0153759350fb5bfe8cd68e3d600516 Author: yihanwu1024 <yihanwu1024> Date: Fri Jun 21 00:00:00 2024 +0000 create post about formal and non-formal judgements diff --git a/6f255651b9b007a6849a55f0ba47de341f034862b6e5ebbe150830128adaf189 b/6f255651b9b007a6849a55f0ba47de341f034862b6e5ebbe150830128adaf189 new file mode 100644 index 0000000..918c837 --- /dev/null +++ b/6f255651b9b007a6849a55f0ba47de341f034862b6e5ebbe150830128adaf189 @@ -0,0 +1,18 @@ +<?xml version="1.0" encoding="utf-8"?> +<section xmlns="http://docbook.org/ns/docbook" xml:id="6f255651b9b007a6849a55f0ba47de341f034862b6e5ebbe150830128adaf189"> +<section lang="zh"> +<title>非形式化的判断</title> +<para>形式化是坚持<emphasis>让人的判断背书自动化判断</emphasis>的一种努力，这是 symbolic AI 的宗旨。不应该反过来断言形式化是它的唯一方法。AI 伦理的要求之一是可解释性，意思或许是 AI 到底只是一个复杂的机器，替人做出等价的推理，而不准产出任何人不能理解的、人的语言不能涵盖的推理。就算放弃伦理标准，这在哲学考验下站得住脚吗？什么才算人的判断背书了自动化判断？只能事后（拿到自动化推理结果）解释，可既然东西不是形式化的，任何推理的事后解释和事前解释也许根本不能相提并论，就算同一人连续做这两步也不一样。</para> +</section> +<section lang="en"> +<title>Non-formal Judgements</title> +<para>Formalization is an effort to, resolutely, <emphasis>back up automated judgements with human judgements</emphasis>. +This is a tenet of symbolic AI. +We shall not conversely assert that formalization is the only way (to back up automated judgements.) +One of the ethic requirements for AI is explainability. +This might mean that AI is simply a complicated machine making equivalent inferences on behalf of humans, while we prohibit what a human cannot understand or what human language cannot convey. +If we skip the ethics part for now, was the above philosophically sound? +What counts as an automated inference backed up by human inference? +It can only be articulated after the fact (after obtaining the automated inference results,) but perhaps if the subject is not formalized, any post-hoc explanations might not be considered comparable to pre-hoc explanations, not even if the same person produced these two explanations consecutively.</para> +</section> +</section> </pre></main></body></html>