In the year since ChatGPT launched to the public, there has been endless speculation about jobs that could be made obsolete by artificial intelligence, but at least one lucrative new skillset has emerged and shown some staying power: prompt engineering.
Google searches for the term — which refers to the art of using keywords to get AI tools to generate better images and written responses — have soared by orders of magnitude from a year ago. LinkedIn said in its November Future of Work report that there have been “substantial increases” in “prompt engineering” and “prompt crafting” appearing on member profiles.
“The joke I’ve heard this year is that the most popular programming language in 2023 is English,” said Justin Farris, senior director of product management at GitLab, a software development platform. He made the remarks during an interview for the latest episode of the Bloomberg Originals series AI IRL, available to stream now.
The rise of prompt engineering positions highlights another way that AI could reshape the job market — not just by eliminating roles but by putting a premium on employees who are adept at using and massaging artificial intelligence services for work tasks. As with all things AI, however, the technology is rapidly evolving in ways that could one day displace these workers, too. Last month, for example, OpenAI rolled out the latest version of its image generator, Dall-E, and suggested it had been improved enough to reduce the need to “learn prompt engineering.
Farris expressed optimism for prompt engineering as a skillset in the near term. ‘In 12 months time, I think more and more people will use it,” he said. “We’re so early. Everyone in tech is talking about it, but it hasn’t sort of crossed the chasm where everyone else is using it.’
Newly-created prompt engineering roles can pay upwards of $335,000 a year, Bloomberg reported in March, and involve people spending their days coaxing AI systems to produce better results or helping companies train their workforces to harness the technology.
“The reason why those folks are commanding such great compensation is that they also possess domain expertise in their field,” Farris said. He compared the skillset to an individual with advanced knowledge of applications like Microsoft Corp.’s Excel, but who also possessed deep understanding of the subject matter behind the numbers and formulae they input into a spreadsheet.
Farris offered several pieces of advice for individuals interested in becoming a prompt engineer or making better use of AI in their jobs. Among them was to learn to be iterative with tools like ChatGPT — and to be patient.
“Have a conversation with one of these systems and pass it more data,” he said. “Expect to get the result you want out of the tenth query, not the first one. Because when you do that, you’re giving the model a lot of context and domain understanding to yield better results down the road. You may not get that on the first query.”
