The world of immediate engineering is fascinating on numerous ranges and there’s no scarcity of intelligent methods to nudge brokers like ChatGPT into producing particular sorts of responses. Strategies like Chain-of-Thought (CoT), Instruction-Primarily based, N-shot, Few-shot, and even methods like Flattery/Position Project are the inspiration behind libraries filled with prompts aiming to fulfill each want.
On this article, I’ll delve into a method that, so far as my analysis reveals, is probably much less explored. Whereas I’ll tentatively label it as “new,” I’ll chorus from calling it “novel.” Given the blistering price of innovation in immediate engineering and the benefit with which new strategies may be developed, it’s completely attainable that this system may exist already in some kind.
The essence of the approach goals to make ChatGPT function in a method that simulates a program. A program, as we all know, includes a sequence of directions usually bundled into features to carry out particular duties. In some methods, this system is an amalgam of Instruction-Primarily based and Position-Primarily based prompting strategies. However in contrast to these approaches, it seeks to make the most of a repeatable and static framework of directions, permitting the output from one operate to tell one other and everything of the interplay to remain inside the boundaries of this system. This modality ought to align effectively with the prompt-completion mechanics in brokers like ChatGPT.
As an example the approach, let’s specify the parameters for a mini-app inside ChatGPT4 designed to operate as an Interactive Innovator’s Workshop. Our mini-app will incorporate the next features and options:
Work on New IdeaExpand on IdeaSummarize IdeaRetrieve IdeasContinue Engaged on Earlier IdeaToken/”Reminiscence” Utilization Statistics
To be clear we won’t be asking ChatGPT to code the mini-app in any particular programming language and we’ll mirror this in our program parameters.