#Fine tuning confusion

1 messages · Page 1 of 1 (latest)

eager mango
#

I am new to OpenAI and I am using the OpenAI API to connect with my WordPress site. (not for Chat or Assitent).
I would like to know if it is possible to fine-tune the results to include some “advice” from me, rather than providing a direct answer.

Here is a simple example:

If I ask for 3 restaurants in Miami, the answer will be a “random from OpenAI” list of restaurants. However, I would like one of the three restaurants to be from my “fine-tuning” dataset.
This way, whenever someone tries to find restaurants in Miami (regardless of the specific question, such as “Show me 3 restaurants in Miami,” “Show me the best things to do in Miami,” or “Make a plan for 2 days in Miami”), the answer will always include that specific restaurant.
I just need to know if it is possible or not.

hot smeltBOT
#

Hi @eager mango. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

#

Based on the information provided in the extracts, fine-tuning an OpenAI model like gpt-3.5-turbo can help refine the way the model responds, but it's not a reliable method for teaching the model new data or ensuring that specific information (like a particular restaurant) is included in the output source (https://community.openai.com/t/314338).

Fine-tuning is more about teaching the model new patterns and ways of handling data. It can be used to set the style, tone, format, or other qualitative aspects of the model's responses, improve reliability at producing a desired output, correct failures to follow complex prompts, handle many edge cases in specific ways, and perform a new skill or task that’s hard to articulate in a prompt source (https://community.openai.com/t/408427).

If you want to ensure that a specific piece of information (like a particular restaurant in Miami) is included in the model's responses, you might want to look into using embeddings and a vector database information retrieval system source (https://community.openai.com/t/328571). This approach could allow you to give the AI relevant information to the user's question.

However, it's important to note that the specific implementation details for this approach are not provided in the extracts. For more detailed guidance, you might want to refer to the OpenAI Platform documentation (https://platform.openai.com/docs/guides/embeddings) and the OpenAI Cookbook (https://cookbook.openai.com/articles/text_comparison_examples#customizing-embeddings).

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon: