Here’s How You Can Use Chat Plugins To Address Shortcomings With ChatGPT & LLMs
LLMs are great at information processing. They are fair to horrible at information retrieval. This is a search problem.
Properly exposing your business logic and data via a web API is one approach to solving this search problem.
Connecting the API to an LLM via a chat plugin is the initial unlock.
That can be taken further when the API is integrated via the LLM’s own API, and the chat interface is bypassed altogether.
This gets very interesting in headless workflows (processes without humans involved).
This isn’t a solution for people looking to run LLMs on private data, but that will be a solved problem in the next generation or two.
In the meantime, creating headless workflows will help businesses scale for pennies on the dollar. That’s amazing.
Related Posts
Products Don't Need To Be Perfect To Buy Them: Why ChatGPT Plus Is Bargain — You are gaining 10x the $20 price for a paid subscription in cost savings, revenue generation, or y ⟵
6 Steps To Write A Chat Plugin For ChatGPT & Microsoft Platforms — This is low dev effort and high value — if your API has valuable business logic and data ⟶
Problems Product Managers Face When Building Products With Large Language Models (LLMs): And How To Counter — Emerging techniques to combat these shortcomings include chain of thought (CoT), few-shot prompting ⟶
