Artificial intelligence tools such as Chat GPT and Dall-E are sparking fears of the technology automating people out of a job, yet like previous waves of innovation the advent of so-called generative AI is also starting to create new kinds of work.
“The good news is that worker displacement from automation has historically been offset by creation of new jobs, and the emergence of new occupations following technological innovations accounts for the vast majority of long-run employment growth,” Goldman Sachs analysts said in a recent report that also forecast a sizable economic boost from AI.
One role now showing up in job listings: “prompt engineer.” The job’s main function is to help train the emerging crop of AI, also known as large language models (LLMs), to deliver more accurate and useful responses to the natural-language queries that people pose in using such bots. More generally, the goal is to make AI smarter and more capable of accomplishing a wide array of professional tasks.
Notably, and unlike many higher-level jobs in tech, working as a prompt engineer doesn’t necessarily require an engineering or coding background. One job listing for a prompt engineer describes the role as an “art” that’s “a hybrid between programming, instructing and teaching.”
Hot new programming language: English
Andrej Karpathy, a founding member of ChatGPT maker OpenAI and former senior director of AI at Tesla, recently tweeted that a prompt engineer can also be thought of “as a kind of LLM psychologist.”
“The hottest new programming language is English,” he tweeted in January after ChatGPT was publicly released, a reference to the fact that LLMs are trained based on prompts written in plain English, rather than specialized computer code.
The good news for job seekers? Some companies are willing to pay big bucks for such jobs, also referred to colloquially by some prompt engineers as an “AI whisperer.”
“We are all amateur prompt engineers, but there is definitely a nuanced understanding to these models,” said Edward Tian, a student at Princeton University who built GPTZero, an app that can detect whether a text was written by a human being or ChatGPT.
For example, LLMs are better at spitting out text in a certain style — say, in the voice of an elementary school student or a comedian — if they are shown an example, Tian explained.
“You’ll get better results if you say to ChatGPT: ‘Here is an example of elementary school writing and then you make the ask,'” he said. “It significantly improves results.”
Prompt engineering is also typically less structured than traditional research experiments, which begin with hypotheses.
“With prompt engineering, no one really knows what the results are going to be, so we try a bunch of things and hopefully the LLM responds in a positive way,” Tian said.
Seeking “creative hacker spirit”
A range of companies and industries are recruiting prompt engineers.
Anthropic, an AI research company and maker of Claude, an AI assistant, is currently seeking a “prompt engineer and librarian,” according to a job posting on the company’s website. The role involves building a library of prompts that get LLMs to accomplish different tasks.
Requirements for the position at the San Francisco company include familiarity with how LLMs work, excellent communication skills and what Anthropic describes as “a creative hacker spirit,” among other qualifications. Basic programming skills and the ability to write small Python programs are also desirable. The pay: Between $175,000 and $335,000 a year.
British law firm Mishcon de Reya is hiring a “GPT legal prompt engineer.” The role will focus on helping the business “increase our understanding of how generative AI can be used within a law firm, including its application to legal practice tasks and wider law firm business tasks,” the job posting states.
Klarity, a company that helps automate contract review, is hiring its own AI whisperer, who will earn between $130,000 and $230,000 a year to fine-tune LLM applications within the company.
Boston Children’s Hospital in Boston is hiring an AI prompt engineer to work on its digital health platform. The desired candidate will have a strong background in both AI and machine learning (a subset of AI), as well health care research experience. The job entails designing AI prompts for LLMs “as they emerge for health care research studies and clinical practice.”
“Super important” skill
To be sure, any job related to an AI chatbot requires a high level of familiarity and understanding of how LLMs work.
“They have to understand how to code, leverage AI models and understand how to talk to them,” Gabor Soter, founder of Generative Nation, a site that educates the public about generative AI, told CBS MoneyWatch.
That said, Soter expects to see a raft new AI jobs.
“Some people underestimate what it takes, but these are front engineers getting hired for hefty salaries,” he said. “I think it’s a skill that’s going to be super important for everyone, and I would highly encourage everybody who is not a data scientist to play around with these models.”