ELI5 the value a “prompt engineer” provides

2 points by amodk ↗ HN
not done much digging, but my naive understanding is that a "prompt engineer" crafts statements which, when used with a user's input, elicits a more accurate/desirable result from whatever model they're interfacing with.

Assuming that's the case, how is that role not BTFO by shifting that responsibility to the model by tuning it better(?) to return the desired response from a naive input, or by requiring the user to craft the higher-quality prompts themselves (probably not desirable.)

The few sources I've come across appear more hand-wavey than anything, would appreciate some technical insight if possible.

3 comments

[ 439 ms ] story [ 402 ms ] thread
Diffusion systems can’t render a great image given just “a dog” as the prompt. Imagine all the different kinds of dog images the user might be expecting.

A good prompt engineer understands the scope of keywords trained into SD and can translate the artistic vision for a finished image into specific keyword constructions.

Try creating a specific visual idea you have in your mind into a jpeg using Diffusion. You’ll see it’s not easy.

Can the user do that? Sure! Just memorize all the critical artists of the last 400 years and their important works. And learn all the specific terminology that yields great results.

Can a computer do it? Sure! Just fill out dozens of form fields to describe your specific intent.

It's more of a sci-fi concept than a real thing
Well most places I have been there is always "that guy/girl" that most people go to for the more obscure Google searches, because they just seem to know how to ask the right questions for the answer you want.

Well this is the same, just for AI.

This is just another extension of living in a world awash with a myriad of answers, the real skill is asking the right question.