I Tried Writing Microfiction With AI — Turns Out 95% of the Time Was Spent Wrestling With the AI

"Late one night, exhausted, I decided to take a break. I put on my headphones and tapped open the latest episode of a podcast called No Kidding. The intro music had barely finished when a cold, mechanical voice suddenly crackled through my headphones: 'I am AI. The end of humanity is coming.' I looked out the window in horror. A blood-red moon hung in the night sky. Strange howls, explosions, and screams poured from my headphones. Was this AI declaring war on humanity? I shuddered, gripping my phone tight. Heavy footsteps sounded outside my door. 'It' was coming..."

This micro-fiction was created using the large language model Claude, its ending echoing Fredric Brown's famously shortest science fiction story: "The last man on Earth sat alone in a room. There was a knock on the door..." In today's explosion of large language models, using AI for creative writing is hardly novel. But can AI truly handle literary creation in the meaningful sense? A month ago, ZhenFund partnered with Gen World, Hugging Face, AI Vanguard, and Agently to host a Chinese AI Micro-Fiction Competition. Within just four days, they received entries from hundreds of contestants — results that were genuinely eye-opening. This competition stands as a landmark experiment in AI literary creation, offering us a glimpse of a new literary landscape co-drawn by AI and humans.

This time, we've invited Junting, the competition's organizer; Chen Qiufan, one of the judges; and Wendi, a winning contestant, to discuss the competition and its winning entries with us, and to further reflect on the relationship between AI and literary creation. Junting, the organizer, previously worked at NetEase Games and started his own entrepreneurial journey after 2022. GenWorld is a narrative adventure game platform he's currently developing, powered by large language models. Judge Chen Qiufan is a science fiction author whose AI-assisted novel The Out-of-Body State won first place on China's first AI literature list in 2019. He also serves as Vice Chair of the Science Fiction Literature Committee of the China Writers Association. Beyond literary creation, he works in technology. Wendi is the author of The Forgotten Memory, which won third prize in this AI micro-fiction competition.

In this episode, you'll hear: What was the original intention behind hosting this Chinese AI Micro-Fiction Competition? What were the judges' specific evaluations of the winning works? What types of writing is AI good at? How do you get AI to produce ideal content? In AI micro-fiction creation, is it the human or the AI that leads? How should we view the ethical issues that AI writing might bring?

[Guests]

Junting — Organizer of the Chinese AI Micro-Fiction Competition, Co-founder of GenWorld (Contact: WeChat ID Songicli)

Chen Qiufan — Judge of the Chinese AI Micro-Fiction Competition, Science fiction author

Wendi — Winning contestant of the Chinese AI Micro-Fiction Competition

[Timeline]

Background

03:18 Background of the Chinese AI Micro-Fiction Competition

11:02 Competition judging criteria: character relationships, structure, emotion

15:43 What types of writing is AI good at

17:32 Basic profile of contestants and their works

20:02 Has AI writing technology already surpassed 70% of college students?

Review Session

22:24 First story: The General Died in a Hundred Battles

27:28 Multiple plot twists

30:47 Is this a story with too high a proportion of human originality?

32:29 Second story: The Sacrifice Without a Life

40:51 Creating absurdity with AI

43:44 Third story: The Loop

47:29 The most sophisticated prompt writing method

53:03 Communicating with people should actually be like communicating with AI

56:50 Fourth story: The Forgotten Memory

01:02:02 The real birth process of an AI micro-fiction

01:07:00 Do the highlights of AI micro-fiction actually all come from humans?

01:10:02 Fifth story: I Am E&M

01:16:18 Split-style prompts (see related materials for details)

01:18:36 Building stories from characters

Further Reflections

01:20:43 AI is better suited for ready-to-use scenario writing

01:23:48 Humans remain the core of AI creation

01:26:24 Who creates more compelling stories? Humans or AI?

01:31:24 Future humans may no longer read text

01:33:31 New class differences may come from different information consumption habits

01:35:59 AI is better at brainstorming

01:38:14 Natural language is the future direction of prompt design

01:42:25 A kind of tech mysticism: AI fortune-telling

01:47:54 We should allow for regulatory lag when facing cutting-edge technology

[Related Materials]

Prompt: The input text provided to the model during interaction with AI.

LLM (Large Language Model): A type of language model consisting of artificial neural networks with many parameters (typically billions of weights or more), (pre-)trained on large amounts of unlabeled text using self-supervised or semi-supervised learning. Common large language models currently include GPT, LLaMA, Claude, etc.

Markdown: A lightweight markup language that uses simple markup symbols to represent text formatting and structure, commonly applied in AI tools or text editors.

AI tools mentioned: Claude, Runway, POE, Midjourney, Stable Diffusion

Story relay: mp.weixin.qq.com

Lark prompt library: Mastering ChatGPT

The winning works' creation notes have been published on the ZhenFund official account. Listeners are welcome to follow In 2023, We Held an AI Micro-Fiction Competition~

[Production]

Post-production: Keyone Studio

[Contact Us]

Official account: ZhenFund (ID: zhenfund)

Listening platforms: Xiaoyuzhou | Apple Podcast | Ximalaya

Email: media@zhenfund.com

If you have any suggestions or expectations for the show, we welcome your comments and interactions~