frontier
There is one important fact to know when using ChatGPT: its knowledge is limited to before September 2021.
Currently some can connect our ChatGPT to the internet, allowing for more friendly message responses.
A separate article will be published to detail how to use these plugins.
Directly above the example, the answer to the example indicates that the knowledge is as of September 2021:
In the process of using ChatGPT, the question answering scenario is one of the most common applications.
Tips for use
1. In order to obtain reliable information, it is recommended to ask questions of basic facts as much as possible.
Individuals often use it to inquire about academic materials, and I feel very reliable through first-hand experience.
2. In the process of communication, in order to improve the efficiency of the answer and be more in line with expectations, it is recommended to add as many qualifiers as possible to clearly inform the model of the specific content you expect. This method can effectively increase the effectiveness and satisfaction of communication.
Let's show it directly with examples:
The first qualifier has only one, so the content of the answer is also more divergent
There are three qualifiers for the second type, which are relatively in line with expectations, if you want to see it, you can pick one in it, hahaha.
OpenAI's API best practices document (click to view the original article) mentions such a best practice:
Instead of just saying what not to do, say what to do instead.
3. In order to obtain better results, it is recommended to avoid raising opinion questions as much as possible.
The current ChatGPT model (GPT-3.5-turbo-16k-0613) has made some progress in some respects.
But personal advice is to keep an open mind and lower your expectations of answers.