The Power Grab of the AI Doomers

AI existential-risk advocates argue that artificial superintelligence could end human life, and some assign that outcome a substantial probability. The essay contrasts this speculative threat with practical cybersecurity risks, noting that AI helped experts find serious PHP vulnerabilities before criminals could exploit them. It traces the movement’s policy logic to philosopher Nick Bostrom’s thought experiment: if dangerous technologies cannot be recalled once developed, preventing them may require ubiquitous surveillance. The author argues that this once-abstract proposal has become a concrete program for monitoring and restricting research.

A Machine Intelligence Research Institute paper outlines 28 ways to verify a ban on frontier AI, including searches of homes and devices, intelligence gathering, polygraphs, employee surveillance, and scrutiny of researchers’ prompts, code, and records. Such measures echo in US and UK proposals: the Sanders-Casar bill would ban superintelligence, pause advanced AI development, establish a powerful federal regulator, and impose prison terms or corporate shutdowns for violations; a British bill promoted by ControlAI would impose a 20-year ban, while a related proposal would authorize emergency shutdowns of models and data centers. The essay’s central warning is that efforts to avert a hypothetical machine dictatorship could empower governments to surveil private intellectual work, constrain speech and innovation, and concentrate control over computing in a global authority. It says media attention to future AI tyranny has obscured these immediate risks to civil liberties.