Prompt Engineering

from AI to ML


An artist, a coder, an enigma, a still to be fully determined skill set but basically someone who carefully crafts prompt that are submitted to AI to obtain the desired response. The response can be text, image or video.



Absolutely not! Some knowledge of a computer language may help in constructing prompts but there is also a view that not being aware of the strict parameters of coding could be an advantage.



Large Language Models are the AI “brain” that a prompt is submitted to. Knowing what they are, how their parameters such as temperature impact the results along with the different use cases for each is useful knowledge but as with programming there is no requirement to have in-depth knowledge.



Shots refer to the amnout of information specific to your query that you provide the AI with. Effectively a measure of the training on your data that you provide. Zero Shot - you don’t provide any additional information. One Shot - you provide one example of the output you are looking for. Few Shot - you provide a few examples of input and output that you are looking for. These different methods from prompt input apply to all Large Language Models from OPenAI, Cohere, anthropic, Meta and those avaible on platforms such as HuggingFace.



While both new roles in the rapidly evolving AI arena we consider them to be different, and complimentary roles. Not a prefect fit analogy but helps with a mental model.

Prompt enginer is to programmer as AI engineer is to DevOps



Not soley focused on prompt engineer, more the overall field of AI and ML, we maintain an AI summary of papers we have read at papers.


developing the art of AI communication