How would I go about finetuning Deep Seek or Llama to do this?
So that instead of a user looking in a manual they just ask the AI questions and it answers.
I think this will need the AI to do 3 things:
- general awareness of general manuals
- access to all the manual related documentation for a specific product (the specific product that we're creating the AI for)
- understanding of all the synonyms etc. that could be used to seek information on an aspect of the product.
How would I go about finetuning the AI to do this? Please give me the exact steps you would use if you were to do it.
(I know that general purpose AI's such as ChatGPT already do this. My focus is slightly different. I want to create AI's that only do one thing, do it very well, and do it with sparse resources [memory/disk space, low compute]).