Creative prompt engineering can lead to some interesting results. And because interesting results leads to clicks which lead to revenue, then engineering interesting prompts will be a task that people will be asked to do.
Prompt engineering will be a skill that people develop and put on their CVs. They will endorse each other for prompt engineering on LinkedIn. And before long someone will likely have “Prompt Engineer” as a job title.
Which means that soon we might also have1:
The 10x Prompt Engineer who can rapidly produce the most interesting outputs because of their knowledge of a wide range of aesthetics, memes, celebrities and an understanding of the strengths and weaknesses of different AI models.
The Prompt Engine Optimiser who will work tirelessly on behalf of Brands to ensure that prompts that mention them produce only positive output. Seeding datasets with openly licensed images containing branded products and positive messages. Helping influencers get their faces into training sets to improve the visual fidelity of outputs. Arcane analysis of multi-dimensional language models to steer a path through alternative phrases. Just a day in the life of a PEO.
Generative algorithm produced some distasteful output? Need to spin the blame onto the user or those responsible for the underlying data? Need to quickly show how some similar, related phrases create something more palatable? You’ll need a Prompt Doctor.
Prompt-sourcing. Not every company can afford to hire the best 10x Prompt Engineers. So why not go out to the crowd to find the best, most creative ideas and phrases?
As platforms move to monetise prompt-driven outputs created by their users, and with predatory prompt-sourcing paying little or nothing to those involved in generating ideas, communities will move to creating their own platforms for sharing prompts.
Prompts of Their Own is a community created and curated website where anyone can safely share prompts featuring their favourite characters, and evaluate them in a range of open, ethically sourced models.
1 Not really, this is a joke.