Original link: https://www.ixiqin.com/2023/03/11/openai-api-uses-three-levels/
With the gradual promotion of ChatGPT, I have seen a large number of OpenAI-based products appear. If it is to be layered, I think it can be divided into three layers:
Connect layer: just using OpenAI’s API, and not doing much functional improvement
For example, the ChatGPT-Feishu I made by myself is actually at this level. It is more about connecting ChatGPT with some ready-made applications, so the difference is not big. Most of you are rolling out some new features in technology.
In my opinion, scrolling at this level is very limited, because there are only ready-made applications and scenarios. Most of the time, what we can do is to better integrate ChatGPT with the existing ecology, but there is no essence changes on.
Prompt Engineering layer: prefabricated prompts to help users ask good questions
When we take a closer look at the usage of ChatGPT on social networks, you will actually find that most people’s usage of ChatGPT is very simple and rude-ask some questions that were asked in the past to search engines. ChatGPT will give you an answer that looks pretty good.
Behind this question is that most people cannot ask a good question that is precise, clear, and easy to understand.
The products at the Prompt Engineering layer can realize the deconstruction of the problem, and disassemble a complex problem into a set of templates + some content that users can understand and input, thereby reducing the difficulty of asking questions.
The application of this layer is more about different scenarios and the optimization of Prompt, so as to achieve more accurate and high-quality returned content, so as to help users solve problems.
Finetune layer: Fine-tune the model to meet application and business scenarios
OpenAI provides the Finetune capability for models. Developers can prepare their own datasets, upload them to OpenAI, and OpenAI will fine-tune the models. Subsequent developers can use the fine-tuned models for automatic completion.
At this level, everyone is concerned with industry domain cognition and clean data, and it returns to the classic AI industry landing scenario: collecting industry data-clear data-training model-practical application. It’s just that OpenAI has reduced the difficulty of this matter. For developers, the entire process can be completed at a lower cost.
If you’re just playing tickets, then I think tiers 1 and 2 are good. But if you plan to do something serious, then the second layer may be the necessary foundation.
This article is transferred from: https://www.ixiqin.com/2023/03/11/openai-api-uses-three-levels/
This site is only for collection, and the copyright belongs to the original author.