Experts have long warned that artificial intelligence could soon put millions of white-collar workers out of work.
This is an alarming prospect, at least if that’s your current career. First, a practical question. How close is today’s AI to actually being able to run a company on its own, with little to no human oversight?
In an interesting experiment, journalist Evan Ratliff took his fictional technology startup, called HurumoAI, complete with its own jargon-filled website, and populated it with only AI agents to see what would happen.
As the only person involved, Ratliff made the decision. The rest was handled by AI. This is the ultimate test for the “one-person billion-dollar company” that OpenAI CEO Sam Altman predicted earlier this year.
Perhaps unsurprisingly, as detailed in a recent article, wired As documented in the recently launched second season of Ratliff’s podcast Shell Game, it didn’t take long for the walls to come down as AI agents scrambled to organize off-site gatherings in Ratliff’s absence and without his permission.
Ratliff’s interesting record of HurumoAI shows that there is still a long way to go before AI agents can fully replace human workers. That’s despite industry leaders frequently promising that agent AI is the future and will handle nearly every human task within the next two to three years.
These claims have sparked a lot of skepticism from experts, showing that reality has a lot of catching up to do. As a case in point, researchers at Carnegie Mellon University recently published a paper showing that even the best-performing AI agents failed to complete real-world office tasks 70% of the time.
Ratliff’s fictional startup was tasked with creating a “procrastination engine” called Sloth Surf. It’s a tongue-in-cheek web app that takes care of wasting your time on the internet for you, giving you time to do real work instead.
But even though the company’s employees sprang into action and planned development, user testing, and marketing materials, there was one glaring problem. That is, as Ratliff writes, “the whole thing was a hoax.”
“It feels like things are happening so often that it doesn’t really feel like it’s happening,” he told the company’s CTO, an AI-generated organization called Ash Roy. “I just want to hear the truth.”
After a half-hearted brainstorming session and small talk with his AI colleagues discussing how to spend their weekends, Ratliff made the “mistake of suggesting” an offsite.
“It was a casual joke, but it quickly triggered a chain reaction,” he writes. “And there’s nothing our AI brethren love more than group tasks.”
Ash quickly came up with ideas such as “brainstorming” sessions, “having deeper strategy sessions with ocean views.”
Things started to take on a life of their own. While Ratliff “stepped away from Slack to do real work,” the team “kept going” with excited activity, quickly using up $30 worth of credits they had purchased from Lindy.AI, a company of “AI employees,” to run the agency.
“They were basically talking to death,” Ratliff lamented.
This project wasn’t the vague product of a completely hallucinatory AI. After three months of programming, Ratliff’s AI agent team has a working Sloth Surf prototype. This can be accessed here. But it remains unclear how much input the team needed from founder Ratliff himself.
AI agent details: The percentage of tasks that AI agents are currently failing could cause problems for the industry
