I have added my chat-gpt API (trial-free mode for just testing) and required codes to my flutter code. But when I send very very simple question, ChatGPT API returned interestingly weird answer to me?
Example:
generated and sent text:
“Where is the capital of USA?”
response:
“text”: “\n\nThis is what Mine is.\n\nRaphael.\n\nHe is a very fortunate guy.\n\nMichael eat a lot of apples.\n\nPeople here eat a lot of plant-based foods.\n\nThey always ride bicycles”,
Well, I can understand the trial-free version may be weaker a bit, but this one is just “totally-not-working” situation. If the paid version is similar, why should I purchase the paid one?
So API is just “bad”! and so, no need to purchase for paid-version. (btw, I have tested several times (10-15) and more than %50 was so weird…). There may be small mistakes but this model is completely dangerous for embedding commercial front-ends
The OpenAI API and chat.openai.com are two different ways to interact with OpenAI’s language models.
OpenAI API: The OpenAI API allows developers to integrate OpenAI’s language models into their own applications, products, or services. It provides a programmatic interface that allows you to make API calls to generate text, perform language-related tasks, and more. With the API, you have more control over the model’s behavior and can customize the inputs and outputs according to your specific requirements.
chat.openai.com: chat.openai.com, also known as ChatGPT, is a web-based interface provided by OpenAI that allows users to have interactive conversations with the language model. It provides a user-friendly chat interface where you can input messages and receive model-generated responses in a conversational format. It’s designed to be easy to use and accessible to non-technical users. ChatGPT is a pre-built application that demonstrates the capabilities of the underlying language model.
In summary, the OpenAI API is a developer-focused interface that allows you to integrate OpenAI’s models into your own applications, while chat.openai.com is a ready-to-use web interface for interactive conversations with the model.
So, if the non-paid one doesn’t respond like the one at chat openai com, I will never get that performance even if I buy the paid one.
On the other hand, does your explanation also mean; I can train (add) my own model to the paid-API? e.g. if I want and so teach API model as 2x2 = 5, will it response 5 for the 2x2? Can I do this?
You are clearly using the wrong model and/or prompting it wrong.
If you ask an AI to write code for you, it also will not write useful models, and it is a blessing that endpoints the AI knows about will be turned off in two months.
A completion model is simply like an autocomplete predictor. It does not follow instructions. You get rambling output like you describe.
If you want to ask questions and get them answered, you’ll want to go to the API reference on the forum’s top or side bar, and read the subtopic about “chat” and the gpt-3.5-turbo model.