#computing #ai
Lately I had to explain further to non-technical people kind of how fine tune works from a usage point of view, not the model details. Here is an example with how OpenAI [[GPT3]] manage fine tuning:
Say you teach a kid to do additions:
You give him some data used for fine tuning by telling him these examples:
1+1=2
2+2=4
Then the kid, understanding these examples, should be able to generalize to:
3+3=?