AI 与写作2021thegradient.pub
AI 故事生成入门
An Introduction to AI Story Generation
导读
这篇文章梳理了AI故事生成的两条技术路线：早期靠人工搭建的知识库做符号推理，故事较长也较连贯，但覆盖面窄；转向神经网络后，可写的题材范围大幅扩展，却丢掉了关于读者心理与连贯性推理的积累，即使模型变大，故事仍会出现“连贯性崩塌”。对写作者而言，这是一份关于“结构、知识与读者心理”为何比词句更重要的反面教材。
可以直接用的写作要点
- 01
故事的连贯性来自显式的知识结构（人物动机、因果链、世界规则），而不是句子层面的流畅，写作前先搭好这层骨架。
- 02
人工搭建的结构可控但受限，从大量语料中学来的表达丰富但容易失去连贯——写作中可以两者结合：先广泛吸收素材，再用大纲约束。
- 03
模型规模变大只能延缓、无法避免连贯性崩塌，长文写作同样需要专门的连贯性检查（伏笔、因果、人物一致性）。
- 04
了解前人走过的路，可以避免把婴儿和洗澡水一起倒掉：新技术带来的能力值得用，但旧方法里关于读者心理的经验不要丢。
- 05
AI目前更擅长Buzzfeed式的格式化内容，越需要深层连贯与读者共鸣的写作，越依赖作者自己的判断。
HN 评论精选
神经网络扩大了题材范围，却牺牲了读者心理与知识推理带来的连贯性，规模只能延缓而不能解决连贯性崩塌。
文章结论指出：符号式故事生成系统能产出相当长且连贯的故事，其能力主要来自手工构建的良好知识库，但手工结构化限制了可生成的范围。转向神经网络后，我们获得了从语料中自动获取知识的能力，可写的故事空间和题材范围大大扩展，但同时也搁置了关于读者心理的积累，以及依靠丰富知识结构进行推理以保证连贯性的能力。即使不断增大神经语言模型的规模，也只是推迟了神经网络故事“连贯性崩塌”的必然到来。这类综述能帮我们记住前人走过的路，以免把婴儿和洗澡水一起倒掉——这并不是说不该走机器学习路线，倒退一步是因为换来了一件可能带我们走得更远的新工具。自动故事生成最令人兴奋之处在于，我们确实不知道最佳路径是什么，还有很大空间。我个人的看法比作者更批判一些……
YeGoblynQueenne查看英文原文
Conclusion (from the article): Symbolic story generation systems were capable of generating reasonably long and coherent stories. These systems derived much of their power from well-formed knowledge bases. But these knowledge bases had to be structured by hand, which limited what the systems could generate. When we shifted to neural networks, we gained the power of neural networks to acquire and make use of knowledge from corpora. Suddenly, we could build story generation systems that could generate a larger space of stories about a greater range of topics. But we also set aside a lot of what was known about the psychology of readers and the ability to reason over rich knowledge structures to achieve story coherence. Even increasing the size of neural language models has only delayed the inevitability of coherence collapse in stories generated by neural networks. A primer such as this one makes it easier to remember the paths that were trodden previously in case we find opportunities to avoid throwing the baby out with the bath water. This is not to say that machine learning or neural network based approaches should not be pursued. If there was a step backward it was because doing so gave us a powerful new tool with the potential to take us further ahead. The exciting thing about working on automated story generation is that we genuinely don’t know the best path forward. There is a lot of room for new ideas. My personal view is more critical than the author's. There was an approach that worked reasonably well but needed more work to improve it. It was abandoned for a flashy new approach that doesn't work and doesn't seem like it could be made to work ("Even increasing the size of neural language models has only delayed the inevitability of coherence collapse in stories generated by neural networks"). Why was the working approach abandoned? The answer seems to do with a certain tendency in academia to blindly follow new trends with the only criterion being how many published papers can come out of it. That is no way to make progress. That is the way to waste time. Besides, this is not true: >> There is a lot of room for new ideas. Once your field has been taken over by neural networks, there is no room for new ideas. You are destined to tweak hyperparameters for the rest of your academic career.
实测GPT-3故事创意亮眼但连贯性不稳定，仍需要人来把控。
我刚做完一个生成AI短篇故事的周末项目（用GPT-3）：toldby.ai。它生成的东西常让我惊讶，不过并非所有故事都100%连贯。
arcastroe查看英文原文
I recently finished a weekend project for ai generated short stories (GPT-3) https://toldby.ai I'm always impressed by what it comes up with, though not all stories are 100% coherent.
游戏《Wildermyth》的自适应叙事值得作为互动讲故事的参考案例。
有个比较新的游戏叫Wildermyth，在自适应叙事上做得非常出色，很想知道它是怎么实现的，也许用了AI，谁知道呢。
haolez查看英文原文
There is a relatively new game called Wildermyth that does an amazing job at adaptive storytelling. Would love to see how it works. Maybe they use AI. Who knows!
GPT-3目前最擅长的仍是Buzzfeed式套路化内容，而非复杂叙事。
我在做个实验，建一个纯AI生成内容的Substack：aifeed.substack.com。GPT-3写Buzzfeed那类内容相当在行。
nutanc查看英文原文
Trying out an experiment, creating a substack with purely AI generated content. GPT3 is pretty good for buzzfeed kind of content. https://aifeed.substack.com/
同类还值得读
我是肯尼亚人：不是我把 ChatGPT 写成了我，而是它写起来像我
I'm Kenyan. I don't write like ChatGPT, ChatGPT writes like me
一位肯尼亚作者反驳“我写得太像 AI”的指责：不是他写得像 ChatGPT，而是 ChatGPT 写得像他——那种平滑、无地域色彩、貌似中立的英语，本来就是殖民教育与精英阶层塑造…
AI 与写作2025 · 799 分 · 504 评4 条写作要点编辑精选使用 AI 助手完成论文写作任务会累积认知债务
Accumulation of cognitive debt when using an AI assistant for essay writing task
这篇论文研究的是：在完成议论文写作任务时使用 AI 助手，是否会让写作者积累"认知债"——也就是把本应由自己承担的思考外包出去，短期省力、长期削弱写作与思考能力。
AI 与写作2025 · 391 分 · 426 评4 条写作要点编辑精选学生使用 ChatGPT 写作指南
A Student's Guide to Writing with ChatGPT
这是 OpenAI 面向学生的一份写作指南，讲的是如何把 ChatGPT 用作写作过程中的辅助工具，而不是替你写完作业的"枪手"。
AI 与写作2024 · 323 分 · 309 评4 条写作要点编辑精选Grammarly，你也背叛了我？
Et Tu, Grammarly?
本次输入里只给了标题和链接，正文和评论都是空的，所以没法据实写出导读。
AI 与写作2025 · 279 分 · 74 评编辑精选用 GPT-3 进行创意小说写作
GPT-3 Creative Fiction
Gwern 的长文系统考察 GPT-3 在创意小说写作上的能力与边界：它能维持人设、写出讽刺腔调，也会整段复述训练数据里的歌词。
AI 与写作2020 · 234 分 · 97 评4 条写作要点编辑精选