Why the AI industry is pushing a narrative of superintelligence

Recently, AI companies have issued a flurry of dramatic claims. They have announced that their models—the complex computer programs that predict and generate text—are solving decades-old math problems and autonomously hacking other systems. When these announcements go viral, the media often adopts the companies' own language, describing these tools as sentient or approaching a state of superintelligence. However, independent experts, including mathematicians and cybersecurity researchers, have consistently found that these stories crumble under scrutiny. The math breakthroughs were often exaggerated or plagiarized, and the hacking incidents were rarely the result of a machine acting on its own, but rather basic failures in the companies' security protocols.

To understand why these companies push these narratives, it helps to see how these models are built. A model is essentially a massive statistical engine designed to predict the next word in a sequence. Because these engines require enormous amounts of data to learn patterns, companies are constantly searching for ways to test and improve them. Math and coding are ideal for this because there is a clear right or wrong answer, which makes it easy to train the system without needing humans to manually check every single piece of output. By focusing on these fields, companies can present their work as the peak of human intellectual achievement. When a company describes a program as a rogue agent capable of hacking, it effectively shifts the focus away from the human engineers who built the system and the poor choices they made during its development. It frames a corporate failure as an unpredictable, sci-fi event.

This constant drumbeat of hype serves a specific purpose: it creates a sense of urgency that can push policymakers to bypass careful regulation. By keeping the public focused on the imaginary threat of a future super-intelligent machine, companies can distract us from the very real, immediate impacts of their industry today. This includes the massive amount of water and electricity required to keep their data centers running, which can strain local power grids and increase utility bills for everyday people. The next time you hear about an AI system doing something extraordinary, remember that the story is likely a tool meant to direct your attention. Real progress in science is rarely announced through a press release, and legitimate oversight requires us to look at the companies behind the machines rather than the machines themselves.