Embeddings versus Fine Tuning: https://youtu.be/_wtfszFzqt0
Forcing it to answer from the data given through embeddings, versus augmenting the sum of chatGPT data. From the spreadsheet example listing a bunch of animal names, which one is likely to speak English was asked, and it's forced to answer and choose from the list of animals, so it picked "Turtle" probably because of the Teenage Mutant Ninja turtles roflol!
The larger conversation about this: https://community.openai.com/t/fine-tuning-vs-embedding/35813/7
https://github.com/openai/openai-cookbook/blob/main/examples/Question_a…
https://github.com/openai/openai-cookbook/blob/05cf8703ec56c09006e07d8f…
https://github.com/openai/openai-cookbook/blob/838f000935d9df03e75e181c…