All GPTs operate through natural language; their goal is to abandon formatting and command-based interaction so that ordinary people can achieve their goals through text input. So this is a simple input-and-output process. In this process, the content we input is called a prompt. Now let us explain the entire human-computer interaction process and the precautions involved.
Comprehension Ability and Characteristics
ChatGPT and Baidu's GPT are different. ChatGPT is like a foreigner in China: although he knows China very well, he does not fully understand it. This understanding is reflected in that it knows your typos and homophones. This incomplete understanding is reflected in local customs, emotional tone, grammar interpretation, and understanding feedback, among other aspects.
Information Reception and Processing
This kind of error in understanding and feedback is reflected in the fact that after it receives your information, if you have not made it understand clearly, it will first change your words into its own understanding, and then execute using the altered sentence.
Prompt Input Techniques
So if you want to use it well, you have to accommodate it a bit. You cannot speak entirely according to your own habits, otherwise it will often misunderstand. This technique is prompt technique.
Control of Output Content
The role of prompts is to control its answers. Sometimes, although GPT understands correctly, the answer may have some problems. When beginners first use GPT, they do not feel this very clearly. When GPT outputs content, different content has different default formats, including Q&A, articles, novels, papers, and so on, but our requirements are usually personalized, so you need to control its output.
The execution process of an untuned GPT:prompt\understanding\understanding deviation\execution\personalized feedback
The execution process of a well-tuned GPT:prompt\understanding\correct understanding\execution\feedback as you want |
|NIUBOYI
( 京ICP备17025393号 )|网站地图