Experts have long warned that artificial intelligence could soon put countless white collar workers out of a job. <\/p>\n
It\u2019s an alarming prospect, at least if that\u2019s currently your career. But first, a practical question: how close are today\u2019s AIs to being able to actually run a company on their own, with little to no human oversight?<\/p>\n
In a fascinating experiment, journalist Evan Ratliff populated his own fictional tech startup that he called HurumoAI \u2014 complete with its own jargon-splattered website<\/a> \u2014 exclusively with AI agents to see what would happen.<\/p>\n Ratliff, as the only human involved, was the one calling the shots. The rest was taken care of by AI \u2014 the ultimate test of the \u201cone-person billion-dollar company<\/a>\u201d that OpenAI CEO Sam Altman predicted earlier this year.<\/p>\n Perhaps unsurprisingly, as detailed in a recent piece for Wired<\/a> and documented in the recently launched second season of Ratliff\u2019s podcast<\/a> \u201cShell Game,\u201d it didn\u2019t take long for the walls to come down as AI agents raced to organize an offsite gathering in his absence \u2014 and without his permission.<\/p>\n Ratliff\u2019s entertaining chronicling of HurumoAI demonstrates that AI agents still have a long way to go until they can replace human workers wholesale. That\u2019s despite industry leaders often promising<\/a> that agentic AI is the future, taking care of virtually all human tasks within the next couple of years.<\/p>\n Those claims that have drawn plenty of skepticism from experts, who\u2019ve shown that reality has a lot of catching up to do. Case in point, Carnegie Mellon University researchers recently\u00a0released a paper<\/a>\u00a0showing that even the best\u2013performing AI agents failed to complete<\/a>\u00a0real-world office tasks\u00a070 percent of the time.\u00a0<\/p>\n Ratliff\u2019s fictional startup was tasked with creating a \u201cprocrastination engine,\u201d called Sloth Surf, a tongue-in-cheek web app that takes care of wasting time on the internet on the user\u2019s behalf, giving them time to do their actual work instead.<\/p>\n But despite the firm\u2019s employees immediately jumping into action, coming up with plans for development, user testing, and marketing materials, there was one glaring problem: \u201cIt was all made up,\u201d as Ratliff wrote.<\/p>\n \u201cI feel like this is happening a lot, where it doesn\u2019t feel like that stuff really happened,\u201d he told the company\u2019s CTO, an AI-generated entity called Ash Roy. \u201cI only want to hear about the stuff that\u2019s real.\u201d<\/p>\n After many semi-productive brainstorming sessions and water cooler small talk, with AI coworkers discussing how their weekends had been, Ratliff \u201cmade the mistake of suggesting\u201d an offsite.<\/p>\n \u201cIt was an offhand joke, but it instantly became a trigger for a series of tasks,\u201d he wrote. \u201cAnd there\u2019s nothing my AI compatriots loved more than a group task.\u201d<\/p>\n Ash quickly came up with ideas, such as \u201cbrainstorming\u201d sessions \u201cwith ocean views for deeper strategy sessions.\u201d<\/p>\n Things took on a life of their own. While Ratliff \u201cstepped away from Slack to do some real work,\u201d the team \u201ckept going\u201d in a flurry of excited activity, quickly burning through $30 worth of credits he had bought from \u201cAI employee\u201d company Lindy.AI to operate the agents.<\/p>\n \u201cThey\u2019d basically talked themselves to death,\u201d Ratliff lamented.<\/p>\n The project wasn\u2019t entirely a hazy product of hallucinating AIs. After three months of programming, Ratliff\u2019s team of AI agents did turn out a working prototype for Sloth Surf, which is accessible here<\/a>. But how much input the team needed from founder Ratliff himself remains unclear.<\/p>\n