How chatbots became AI’s biggest talking point

When the World Economic Forum predicted a few years ago that artificial intelligence would cause seven million job losses, the great and the good of Davos were able to wring their hands about the human cost of progress while feeling certain that they would be left unscathed. Automation might, they believed, take out office drones and administrative roles but not management gurus or software engineers.

That comfy confidence has been shaken in recent months by a new wave of “generative AI” tools. Apps that can create pictures, video or prose good enough to pass as authentically human. At the moment one in particular has captivated Silicon Valley. Hailed as the smartest chatbot ever made, Chat GPT can generate high-school essays, software code or marketing strategies within seconds, all from a few words of a “prompt”. Social media timelines have been filling up with people trying out the technology. “This used to be my job,” tweeted one former Meta manager, after ChatGPT had churned out a plausible take on Instagram’s product road map.

ChatGPT’s results are not always reliable or accurate. Nonetheless, all of a sudden, AI has the creative classes and middle-management squarely in its sights. But fear not, Davos-goers. The WEF predicted that AI would not only kill jobs but create them. And one such role that has emerged with the rise of generative AI is the “prompt engineer”. 

This job description sees writing the prompts necessary to get the best responses from AIs as a skill all of its own. Wrangling ChatGPT requires a deeper understanding of how AI works — hence “engineer” — as well as domain expertise, be that coding, marketing or homework. An online marketplace called Promptbase has already sprung up, where prompt engineers can sell their carefully crafted instructions for image-generation tools such as Midjourney. For a couple of bucks, you can buy pre-written templates for “cute robotic animal pictures” and “3D game renders”. 

AI entrepreneur Colin Treseler, co-founder of Supernormal, is one of those looking to hire. “You have to find ways to talk to the model to get the correct output,” he says. The problem is there aren’t many people doing this job yet: I found only a handful on LinkedIn.

On first mention, prompt engineering sounds like a comforting bedtime story for knowledge workers, in which preparing for the future involves little more than messing about with ChatGPT and posting the results on Twitter. Lots of the tech elite have spent the last couple of weeks doing little else anyway. But Basil Safwat, design lead at AI start-up Adept, thinks I am being a bit too cynical. “There is genuinely a need for people who are slightly ahead in their understanding of this new material,” he says, which he says is still in its “raw state” today.

Indeed, the very notion of a prompt engineer reveals the biggest shortcoming of these new AIs: they aren’t quite smart enough yet for just anyone to be able to use them successfully. And things are moving fast. Safwat believes that soon the interfaces we use to access and manipulate these AIs will improve, in the process making prompt engineers redundant. Likening it to the evolution of the computer operating system from a text-driven command line to windows and touchscreens, he says: “I don’t think this stage will last for long.” 

Perhaps what prompt engineers really represent is a whole new class of employment disruption: jobs that have been both created and then destroyed by AI.

Tim Bradshaw is the FT’s global tech correspondent

Follow @FTMag on Twitter to find out about our latest stories first