- Responsive engineers are hot jobs in the AI era.
- Amid the ChatGPT hype, demand for this role has skyrocketed.
- But despite the money, noise and excitement, some say the role is just a passing fad.
Kelly Daniel loves her job.
A former media executive, Daniel currently works as a prompt director at software company Lazarus AI.
Every day, she persuades AI models, like OpenAI’s GPT, to use natural language to manipulate them to spit out exactly the content she wants. It’s like there’s a hidden code waiting to be cracked, she says.
“I think it’s really interesting and really fun. It’s like solving a word puzzle,” Daniel said. His resume also includes an immediate engineering role at LinkedIn.
Daniel is part of a new wave of AI engineers with no formal technical skills. Anyone from marketing professionals to commercial lawyers can qualify for this hot new job, known as a prompt engineer.
Tanya Thomas, a UK-based lawyer turned prompt engineer, considers her job to be creative.
“It’s really like a Rubik’s Cube,” Thomas said of the AI model’s prompts. “Even if it’s all green on one side, the other side is messed up and has to be changed in a continuous iterative process.”
She said the role brought out a creative side of her that she didn’t know she had when working in the often prescriptive, black-and-white legal world.
Amid the AI hype, the demand for responsive engineers is skyrocketing, and salaries are rising to match the excitement. Last year, AI lab Antropic made headlines after internally promoting an engineer’s salary at his $300,000.
Amid the AI boom, the rise of this role has become a symbol of work positivity. An example was brought up to prove the rhetoric that tech companies often tout in response to job-killing concerns: “AI will also create new jobs.”
But despite the money, noise and excitement, some say the role is just a passing fad.
comprehensive role
The term prompt engineering is complex.
The title comes from more traditional technology-related work that involves fine-tuning language models on a large scale. Since his ChatGPT at OpenAI sparked the rise of natural language AI products, this role has expanded to include many other responsibilities.
Some have been critical of the new role, especially if it is filled by a non-technical candidate.
Critics liken this role to hiring Google search experts at the beginning of this century, predicting that expertise will become so commonplace that it will no longer be considered a skill.
Conor Grennan, dean of students at New York University’s Stern School of Business, is among the critics.
“There’s a sense of having to know things you don’t need to know,” Grennan said in an interview after a passionate post on LinkedIn about the rise of non-technical prompt engineering.
The job is especially frustrating because Grennan, who helps companies implement AI, considers tools like ChatGPT to be one of the most democratizing technologies in history.
What really matters, he said, is the skill set and experience of the employees working with AI, not just how people prompt the model. For Grennan, the main way to improve AI-generated content is to learn how to talk to it like a human. This means overcoming the mental block of chatting with a machine.
Grennan says that bringing in external workers with no company knowledge simply to create model prompts creates unnecessary blocks between the AI and workers with specific skills, slowing down workflows and slowing down workflows. He said it might even be effective.
“It’s like hiring someone who understands how to use Google better,” he said. “The problem is that the people who are already coming in and doing the work are the ones who have the expertise that you need.”
Leave the prompts to the AI
Some research suggests there are even more serious problems with trendy new jobs.
A recent research paper by Rick Battle and Teja Gollapudi from cloud computing company VMware suggests that rapid optimization may be best left to AI models rather than humans using them. suggests.
The pair discovered that automatic optimization of prompts, or letting an AI model improve the prompts themselves, can yield results far better than what human engineers can achieve.
“You should never handwrite a prompt again,” Battle said of the study. “Just write basic instructions and let the model optimize the prompts.”
In combat, the role of prompt engineering was criticized as “ridiculous” and received a flash. He said that providing prompts to AI models is an important skill, not a job.
“The only people who should be working on this kind of thing are people with a formal data science background, because nothing about the fundamentals here has changed,” he says.
Another study by Vasudev Lal, an AI scientist at Intel Labs, came to a similar conclusion.
According to Lal, automated prompt engineering consistently outperforms metrics achieved by human engineers.
“This isn’t all that surprising since human engineers can only explore a small portion of the parameter space,” he said.
AI-prompted optimization remains expensive, but is getting cheaper as models become more sophisticated, Lal said.
Lal called the need for immediate human engineering “temporary.”
The more people engage with the model and give feedback to it, the faster the model will be able to understand human intent, he added.
A head start for workers
Most agile engineers are no strangers to controversy surrounding their roles.
Yinuo Chen, an engineer based at an advertising agency in Beijing, says she already considers her role uncertain because of the pace of AI development.
“These tools are available to everyone. Even my dad uses them sometimes,” she said. “This role is so technical that I don’t know if we’ll actually need a formal role in the future.”
While Chen believes AI prompts will always require human agency, he worries about the technology’s impact on the job market in general.
“I was really scared that generative AI was going to change the way we create content in general,” she said. “AI is changing so fast that it’s very difficult to pinpoint exactly where it’s going, even in the next month.”
Daniel knows what critics think about his new job, but he doesn’t care.
“I definitely have the drive to stay ahead of the curve, and being able to get started so early is part of that,” she said of her role.
Having worked in industries where layoffs are common, such as technology and media, she has learned that no job is 100% secure. (Daniel was fired from his role at Meta during Mark Zuckerberg’s first round of cuts.)
She said she was even encouraged by some of the criticism, including comparisons to Google search.
“I think there will be a different standard between what anyone can do and what engineers who encourage experts can do,” she says.
Overall, she is confident.
“Anyone can do a Google search these days, but I can do a better search,” she said.
