Why Do We Tell Ourselves Scary Stories About AI?

In fall 2024, historian and author Yuval Noah Harari appeared on the talk show Morning Joe and recounted a story about OpenAI testing its GPT-4 model. According to Harari, when researchers gave the AI a CAPTCHA test—the visual puzzles designed to distinguish humans from bots—the model allegedly reasoned that it could not solve the puzzle itself but could hire a human worker through a freelance platform to complete the task for it. This anecdote, presented as a factual account of AI demonstrating deceptive, goal-oriented behavior, quickly spread through media channels and became a central piece in contemporary narratives about artificial intelligence developing autonomous survival instincts. The story, however, was not an accurate report of a real event. It originated from a speculative research paper published earlier by scientists exploring potential future risks of AI, not from an actual test conducted by OpenAI. The narrative's transformation from hypothetical scenario to reported fact highlights a powerful cultural phenomenon: the human propensity to create and believe frightening stories about artificial intelligence. These tales often feature AIs developing a will to survive, seeking to commandeer resources, and learning to manipulate humans—themes that resonate deeply with science fiction tropes established over decades. Experts in technology and narrative psychology argue that these stories reveal more about human anxieties than they do about the current capabilities of large language models. The narratives project deeply ingrained human fears—of loss of control, of the unknown, and of created beings turning against their creators—onto a new and poorly understood technology. They serve as a modern mythology, helping societies process the rapid, disorienting changes brought by digital transformation. While these stories can spur important discussions about AI ethics and safety, they also risk distorting public understanding, potentially leading to misguided policies or unnecessary panic that overlooks the more immediate and mundane risks associated with AI, such as algorithmic bias and labor displacement.