Prompt engineering can potentially pay up to $335,000 amid the latest trend in work shift. And you don’t even need a CS or math degree.
Latest in work news influenced by AI is the high-paying job of a prompt engineer. A recent job listing showed that companies are willing to pay as much as $335k for a prompt engineer.
There are many questions surrounding prompt engineering’s proliferation. Is it even here to stay? Are dedicated prompt engineers going to win in the long run vs. those who incorporate AI generators and LLM tools in their current workflow? Will more advanced AI tools even need prompt engineers, as they become even better at understanding our intent?
The answers to those questions, like many others involving AI popular culture today, are not concrete. Yet, this space is booming for now. In this case study, we’re going to look at 5 important factors:
What & why?
A prompt engineer is a person who can write and fine-tune prompts for ChatGPT (and other chatbots and AI image generators) to come up with the right responses for the job. ChatGPT allows anyone to write a message and get a response. But it can be manipulated, coerced, or forced to give responses in a particular manner which is the most beneficial for the job at hand.
Major tech companies are rolling out their AI tools. Google (through DeepMind, Bard), Microsoft (through OpenAI and Bing), Meta, Baidu (through Ernie), etc. are all in the fray.
LLMs have seen a surge in development in 2023. And to interact with almost all of these LLMs, you need prompts. Though these LLMs are pre-trained, you can condition them further for a specific use case. And since they save the histories (most notably ChatGPT), you can pick up where you left off, seemingly creating an unending supply of text and images for a particular use case once instructed properly.
Digital marketing specialist Amit Biwaal is a self-proclaimed prompt engineer who, on his website, identifies 3 core areas for development around prompt engineering: control over the output, improved accuracy, and enhanced creativity; though he clearly specifies that a background in the technical aspect is a prerequisite.
A lot of people want to jump in. But you must understand what it entails. Given the competition, it’s probably going to be hard to get a good full-time job as a prompt engineer, but that might soon change.
- In a piece on Bloomberg, a prompt engineer from Accenture, Albert Phelps says that “It’s like an AI whisperer.” He, with his colleagues, spends most of the day writing and fine-tuning prompts and saving them as templates in ChatGPT for the use of the firm’s clients.
- A job listing for a Prompt Engineer and Librarian from Google-backed AI research firm Anthropic has a salary range of $175k – $335k with equity and benefits.
- You can buy and sell prompts for DALL-E, ChatGPT, Midjourney, and Stable Diffusion on a dedicated prompt marketplace called PromptBase.
You can also find a lot of help online. Though a lot of training can be gathered just by watching a couple of YouTube channels religiously, there’s definitely more well-documented work online to make use of.
For example, there is:
- A course for generative AI that teaches text and image prompt engineering.
- A GPT-4 based natural language called Jargon that makes it easy to translate text into LLM-friendly interpretation.
- A well-documented guide on prompt engineering that covers LLaMA, ChatGPT, and GPT-4 (also on GitHub)
- A repository focused on marketing and productivity applications such as creating a social media calendar that prompt engineers can use and learn from.
Doing it right
Back in March, 2021, an AI expert Shubham Saboo wrote in a Medium post that with the no-code revolution just around the corner, prompt engineering and design is the first and foremost thing to consider while creating any GPT-3 (then recent) application, and goes on to provide a 5-step formula for effective prompt training.
That has come true. And doing it right is now a responsibility of those who are trying to make it big as prompt engineers, regardless of their background.
To understand further, we must first learn how the whole process of writing a prompt, getting a less-than-satisfactory response, and the further refinement of the original prompt or the addition of further instructions works.
A working paper [PDF] from OpenAI, OpenResearch, and the University of Pennnsylvania notes that, “Iterating on the prompt, based on observed outcomes in a small validation set, can enhance the agreement between model outputs and the rubric’s intent.”
This enhancement is what one is after during the whole process of prompt engineering.
According to a Twitter handle AI Daily, here’s how to get started:
Understanding the basics will not take you long. Figuring out how to coerce ChatGPT or Midjourney, for example, to give you results that you want is an experimentative hobby of many.
For those seeking a more advanced training, Learn Prompting is a free and open-source course on how to communicate with artificial intelligence (more precisely, the currently popular AI chatbots like ChatGPT and Bard).
From the Financial Times titling their take “Is becoming a ‘prompt engineer’ the way to save your job from AI?” to VentureBeat calling it one of the critical careers of the future, almost every major publication has covered prompt engineering, if only to provide tips or resources.
- Euronews reports that the job involves communicating with the algorithm in order to gradually teach it how to respond while following specific guidelines.
- In an episode on IdeaFlow from MIT’s Computational Law Report, guests explored prompt engineering for a legal brief using ChatGPT in a 1-hour long deliberation.
- In a meticulously argued opinion piece on The Guardian titled “The stupidity of AI,” James Bridle notes how the technique of framing a question such that it’s most clearly understood by the system or how to structure inputs was already something that tech commentators predicted would become a well remunerated job in a no-code future.
- In a late-2022 BBC article, the MIT-trained roboticist and artist Alexander Reben (of the AI Am I series of artworks) narrates how he spent a month on prompt engineering to write more effective inputs for AI systems and how he discovered “a sequence of initial words that would ‘tickle’ the AI in the right way,” ultimately creating a workflow to produce descriptions as he intended.
- Wired has summarized 11 tips that take your ChatGPT prompts to the next level including acting-out a roleplay.
- Marcel Scharth on Tech Xplore/The Conversation, while covering perfecting prompt writing for AI generators, duly noted that there is a duality to it all where some believe that existing writers, programmers, and designers need to incorporate AI generators to stay ahead of the curve vs. those who strongly believe that prompt engineering will become a job of its own, which we now see materializing.
Where do we go from here?
According to Google Trends, the spike in the search query “prompt engineering” approximately follows the uptrend in the growth of the term “ai jobs.”
Prompt engineering is very likely the next popular culture megatrend surrounding AI after the release of AI tools themselves that generated text and images before the time of ChatGPT, up until the release of GPT-4, Bard, Bing’s integration with AI, and other companies launching their own AI chatbots or tools.
Prompt engineering is fast becoming a legitimate job profile where salaries can go as high as those for experienced developers.
It’s definitely a trend in the making. Though it’s nowhere close to the number of AI startups being launched every week, the speed is picking up for prompt engineering tools, job listings, and guides or resources to the point that it’s looking like a boom, as one happened in NFTs, cryptocurrencies, and blockchain space back in 2021.
If you wish to become a prompt engineer, then there is no better time. But you must first do more research than a casual prompt writer and preferably learn more about the inner workings of language models. Whether it’s a gig worth investing time and effort in is completely up to you and depends on the opportunities in front of you.