Over the last few months, there has been increased focus on the role of ‘prompt engineers’ in running LLM applications, but an OpenAI employee says that there might not be much to the hype.
“you should not become a prompt engineer, even if someone paid you to be one,” tweeted Logan Kilpatrick, who works at a Dev Rel at OpenAI. “I’ll caveat all of this by saying prompt engineering is an emerging field, but right now it lacks the foundation for long term success. If you are looking at jobs and think the role of “prompt engineer” is a safe bet, I hope to change your mind,” he continued.
“Many people are looking at AI, thinking about how it will disrupt the job market, and trying to position themselves well for the future. This is 100% the right approach. There’s been a lot of media narrative around the fact that prompt engineering will be the best future job,” he continues.
But Logan says that like other jobs, even Prompt Engineering jobs will likely be taken over by AI. “The problem is that more and more prompt engineering will be done by AI systems themselves. I have already seen a bunch of great examples of this in production today. And it’s only going to get better,” he said.
“Just wait for the day when ChatGPT can synthesize your previous conversations and do some auto prompt engineering for you on your queries based on the chi text that it has. All of this is to say, it’s not clear prompt engineering is differentiated long term,” he added.
“This isn’t to say that people who deeply understand how to use these systems aren’t going to be valuable, but I imagine this will be a skill that is learned as part of people’s standard educational path, not some special talent only a few have (like it is today),” he says.
It’s an interesting argument, especially because it comes from someone who’s from OpenAI, and has a ringside view to how the AI field is shaping up. There have been several reports of job openings popping up that pay as much as $350,000 to ‘prompt engineers’, and people have been looking to learn to to prompt LLMs. There seems to be some value to this skill at face value — the prompts can dramatically change the outputs of LLMs, particularly for image generation for Stable Diffusion and Midjourney. Also, there appears to be a natural correlation between technologies that existed previously — computer engineers were experts who were able to communicate with engineers, and similarly prompt engineers could interact with LLMs. But an OpenAI insider seems to believe that prompt engineering too will eventually abstracted away by the relentless march of AI, and will ultimately chiefly be done by computers.