Highlights from Patterns of Application Development Using AI by Obie Fernandez
Highlights from this book
-
In contrast to the way that you and I think, an AI model’s “thinking” via inference happens in all in one stateless operation. That is, it’s thinking is limited to its generation process. It literally has to think out loud, as if I asked you a question and only accepted a response from you in “stream of consciousness” style.
-
I like to think of my AI components as little, almost-human virtual “workers” that can be seamlessly integrated into my application logic to perform specific tasks or make complex decisions. The idea is to purposely humanize the LLM’s capabilities, so that nobody gets too excited and assigns them magical qualities that they do not possess.