top of page

【Prompt Engineering 101 提示工程基本教學】


AI has been a hot topic for a while now, but have you mastered the right way to generate your prompts? If you frequently use ChatGPT (or other AI tools), keep reading to find out how you can boost productivity with the right prompts! ➡️




Prompts can be zero shot, few shot, or chain of thought. Just like in 🏀, the more you throw, the better your chances of scoring. The same logic applies for ChatGPT - the more prompts you generate, the closer you’ll get to the desired outcome. But, if you do throw the first ball with some technique, it’ll be easier for you to “score”!



A good & precise first prompt should be 200-300 words long. Since your AI doesn't know your context, needs or wants, make sure you've included these elements:

- 👩🏻‍💻Persona: What role do you want the AI to be? Use language like “You are an expert …copywriter”.

- 📄Instruction and output: What is your deliverable?

- 🌁Context: What is your aim, background and audience?

- 👔Style and Tone: How formal do you need the tone to be, or other specific instructions e.g. UK/US spelling?

- 💯Model Answer (s)/ Desired format: Do you have an example? Tell the AI to analyse it and use your voice!

Now that you know the basics, put theory into practice and let us know what you think!


呢排大家一直討論AI呢個hot topic,但係對於有效咁出提示畀AI嘅技巧你又識幾多呢?你哋對於基本prompt engineering(提示工程)又有幾理解呢?如果你經常使用ChatGPT或類似工具,就繼續讀落去啦!➡️


提示主要分幾種,有零樣本提示(zero-shot),少量樣本提示(few-shot)同埋關聯思考提示

(chain-of-thought)。好似打籃球咁,對住個網射得越多次,個波就越大機會入網。正如出提示畀個AI咁,你出得越多,就越接近你想要嘅結果。但係有技巧咁射第一球,就當然更加容易有「分」!


一個準確嘅提示首先需要大概200-300字,將AI係唔知道你背景同需求嘅情況下有結構咁列曬出黎。一個準確直接嘅提示包括咗呢啲元素:

- 👩🏻‍💻形象:AI應該扮演咩角色?用「你是...專家的撰文員」呢種文字去建立一個AI需要代入嘅身分

- 📄指引及輸出:你希望可以透過AI做啲咩? 例如IG post,論文,主持稿等等

- 🌁背景:你嘅目的、背景同受眾係邊個?

- 👔風格及語調:你需要變種語調?例如書面,口語? 加上其他特別要求,例如英式/美式英文。

- 💯理想答案/格式:如果你有具體嘅例子,可以分享畀嘅AI睇,叫佢幫你分析同用返相近語調啊!


依家你有一啲提示工程嘅基本知識,記得將呢啲理論實踐,有咩諗法都可以分享畀我哋聽啦!

Comments


bottom of page