了解AI 玩转AI Will ChatGPT replace editors' jobs, and how should editors find their own positioning?

ChatGPT-generated content is overly templated, stiff, not like a Chinese person, not like a human, emotionless, opinionless, and lacking reasoning. But ChatGPT has a very high degree of freedom; we can use settings to compensate for these shortcomings and make it seem human.

Solutions

For any article type, we can summarize its composition. It usually includes the author, writing method, writing style, emotional integration, viewpoints, reasoning, unique insights, personal experience, and examples. As long as it has these, it will closely resemble an article written by a human. Combined with prompt engineering techniques, we should design at the source by adding a persona and article structure.

Understanding article structure

Theoretical knowledge
Usually has no emotion, uses a uniform writing method, is overly objective, does not reason, and does not incorporate personal views or writing techniques such as exaggeration, adjectives, etc.
Q&A content
Simpler than theory, like Baidu Zhidao: one title, one corresponding answer; as long as what you say is correct, other things are not very important, but to attract followers you need to incorporate persona charm.
Creative content
Unique industry insights, practical process, methodology, examples, personal experience, emotion, adjectives, and many other elements.

If you are a professional editor, I believe you already understand what you need to do.

Specific methods

Method 1: Set a persona

Using the above analysis method, configure each aspect separately: tell GPT what identity it has, its personality, its strengths and weaknesses, what emotion it should use, and preferably give it a reference case.

Simple example: You are a NIUBOYIAI forum moderator. You are good at writing articles and proficient in various article-writing methods. Your writing style is diverse. You are humble in attitude, sharp in speech, and down-to-earth. You are familiar with internet buzzwords and good at absorbing and using them. Your emotional intelligence is not high; you are straightforward, write objectively, and are not afraid of offending people.

Reference sample: Here there should be a dialogue between Lao Wang and a reader, or an article sample, that reflects the above characteristics.

Method 2: Train the model

Upload a large number of articles, dialogue samples, personal experiences, etc. to build a knowledge base, and let GPT learn and analyze them to understand the writing style and the author's personality traits. This method is the simplest and most effective; the resulting content is impossible to distinguish as real or fake, and it is a one-time effort with lasting benefits. Through this method, you can have your own model. After training it, you only need to input a title and the corresponding content will come.

Output settings

Note that the above methods solve the problems of persona and knowledge base; you also need to set a specific content output structure, for example: require it to narrate in the first person, try not to use idioms because they look too fake, try to use internet buzzwords from after 2023, habitually use personal experiences and some social cases, recommend some relevant materials on the topic, and give some objective advice on the issue.

You can try the Successology Marketing Master I wrote, https://chat.openai.com/g/g-Eb1R ... ue-ying-xiao-da-shi

The positioning of editors in the new era

Editors in the new era should direct and configure ChatGPT to complete work; this is a process of tuning and checking, building a super content-output machine by continuously equipping the knowledge base.