我们可以也应该赋予大型语言模型一种奇思妙想的感觉吗？
我在周末花了很多时间深入研究我的claude.ai Max计划的使用方式，以便claude能够以一种与我的个人品味和倾向高度契合的方式创建系统和文档。<p>作为一个有一个相当自闭症孩子的父亲，同时我自己也可能有些在自闭症谱系上，我与孩子们的沟通中充满了荒诞的幽默。<p>有没有人认真思考过，我们是否能够让大型语言模型（LLMs）像维克多·路易斯-史密斯（Victor Lewis-Smith）或类似的讽刺作家一样幽默和令人不安？
查看原文
I spent the weekend really nailing down my usage of my claude.ai Max plan such that claude creates systems and documentation in a way strongly aligning with my own tastes and inclinations.<p>And as a father of one quite autistic child, and likely being somewhat on the spectrum myself, my communication with the children involves a lot of absurd humor.<p>Has anyone thought really hard about whether we can make LLMs as funny and disturbing as, say, a Victor Lewis-Smith, or similar satirists?