[基础使用] What is GPT, and how can one simply understand the relationship between GPT and ChatGPT?

GPT is a generative model for processing data. It both learns a large amount of knowledge and has diverse ways of processing and generating.

Its state is like a super brain; it has learned roughly 70% of human knowledge. How to use this super brain, what scenarios to place it in, and how to define its operating mode all depend on how developers configure it.

When Openai was considering how to use their GPT model, they created the product Chatgpt. Its configuration allows users to run the GPT model through natural language, the simplest input method. This way of using it intuitively demonstrates GPT's data processing capabilities, including receiving information, understanding information, retrieving from the knowledge base, and generating responses.



Chatgpt is suitable for knowledge Q&A and simple task processing; this is its product positioning. For more complex tasks, Openai offers other product lines, such as the API, which can be integrated into certain parts of external applications to achieve automation and unlimited input and output.

Considering the business scenarios of enterprise users, Openai also offers fine-tuned large models, which can train on their own private data based on the technical framework of the base model and define processing methods, thereby enabling business operations in special scenarios.

For small teams, they also offer Gpts for business scenarios with small data volumes, along with a corresponding assistant API for external calls.

One point must be clear: the input and output of all applications built on the base model are determined by the developer's configuration. Chatgpt's configuration is excellent; it can clearly understand our input and the corresponding output.

If you do not configure anything and directly call the APIs of models such as GPT-3.5 and 4.0, using them directly will not produce the same performance as Chatgpt, because you are calling the base model.

In addition to configuration, there are many enhancement and auxiliary modules, such as RAG technology, which is used to retrieve data and is almost essential in the data processing steps. When calling external knowledge, RAG technology can provide auxiliary data from a massive knowledge base for the base model's reference. If the base model is responsible for generation, RAG is responsible for retrieval and providing information.

Only after clearly understanding the concepts can we better utilize it. Only after understanding Openai's product lines and positioning can we choose suitable products to integrate into the work environment.

To learn Chatgpt systematically, please see this article. https://www.niuboyi.com/peixun/