AI researcher Gary Marcus has dismissed the viral essay "Something Big is Happening" by entrepreneur Matt Shumer as alarmist hype, arguing that the claims about AI rapidly replacing jobs are unfounded and lack empirical support. Marcus, known for his critical perspective on AI development, told Business Insider that the essay, which has been viewed over 80 million times on X, presents a distorted picture of AI’s current capabilities. He criticized Shumer’s assertion that AI will soon disrupt nearly all screen-based jobs, calling it a case of "weaponized hype" that exaggerates the technology’s readiness. Marcus emphasized that while AI can assist with certain tasks and improve efficiency, it is far from being able to replicate the full scope of human work, especially in complex or high-stakes domains. “AI can do a small subset of the tasks, and that sometimes speeds up human beings and things like that, but it rarely does all of what a human being can do in any particular domain,” Marcus said. He acknowledged that AI may eventually replace most labor over the long term—possibly within a century—but stressed that the next one to two years will not see mass job displacement. Marcus pointed to real-world examples of companies overestimating AI’s capabilities. He cited Klarna, the Swedish fintech, which in 2024 claimed an AI assistant could do the work of 700 employees. By May 2025, CEO Sebastian Siemiatkowski admitted the company had to scale back and begin hiring humans again, as the AI fell short of expectations. “Six months or a year later, they come back with their tails between their legs because it turns out that the AI systems don't do things as well as the human,” Marcus said. He warned that the real short-term threat isn’t AI replacing junior workers, but rather executives making costly decisions based on inflated expectations. “The biggest thing I think junior people have to worry about right now is a misapprehension by the C-suite that these techniques work better than they actually do,” he said. Marcus also highlighted the lack of solid data in Shumer’s essay, particularly its use of a widely circulated Model Evaluation & Threat Research graph that he argues lacks proper context. He referenced a June 2025 study from Apple’s Machine Learning Research Group that underscores the limitations of current AI models. He noted that many high-profile AI figures have a history of overpromising. Elon Musk’s prediction of a million robotaxis by 2020 failed to materialize, and Geoffrey Hinton’s 2016 call for stopping radiologist training was later clarified as a misstatement, though he admitted the broader trend toward AI in medicine was valid. While Microsoft AI CEO Mustafa Suleyman recently predicted that most white-collar jobs could be automated within 18 months, Marcus remains skeptical—especially in fields like accounting, where accuracy is paramount. “Even one mistake can cost a client hundreds of thousands of dollars or get them sent to jail or whatever. Accounting is a business that is built on accuracy. If you're not accurate, you don't have a business,” he said. In sum, Marcus urged a more measured, fact-based approach to AI discourse, warning that unchecked hype could lead to economic missteps and misplaced confidence in the technology’s current capabilities.