My AI costs went from $100 to less than $1/day: Fine-tuning Mixtral with GPT4

Re:My AI costs went from $100 to less than $1/day: Fine-tuni... by nextworddev on Thursday January 18, 2024 @08:08PM (#39050361)

And even this analysis is optimistic as it doesn’t factor in the $$$$ it costs to hire a data scientist to fine tune the models. Just use off the shelf models with RAG until you really need custom models

[ Parent | on HN ]

Re:My AI costs went from $100 to less than $1/day: Fine-tuni... by ignoramous on Friday January 19, 2024 @07:00AM (#39054459)

You've a point. Doesn't look like RAG (w/ in-context learning) or fine-tuning add domain knowledge to the LLM, so some claim they're equivalent: https://twitter.com/Shahules786/status/1748059074556760421 / https://archive.is/iRN5j

Proxy-tuning (https://twitter.com/rasbt/status/1748021765790376385 / https://archive.is/oQs0m) and other such merged models(https://twitter.com/osanseviero/status/1745121420353454219 / https://archive.is/hFYbh) are proving to be interesting area of study, too

[ Parent | on HN ]