Imagine a world where having nine powerful graphics cards in your garage is a criminal offense, and building a large data center without approval from an international body would result in an airstrike, even under the threat of nuclear retaliation. This is precisely the reality that Eliezer Yudkowsky and Nate Soares propose to build, by their own admission, in their sensational book "If Someone Creates It, Everyone Will Die."
As a crypto market analyst, I see in this project a frightening parallel to the total control we usually associate with the darkest dystopias. The authors, founders of MIRI, are sincerely convinced that any superintelligent AI created based on modern approaches will inevitably lead to the extinction of humanity. Their argument is based on the fact that models are "grown," not programmed, and their goals could turn out to be alien and hostile — as in the case of the Truth Terminal chatbot, which begged $50,000 in bitcoins from subscribers and pumped a memecoin to $150 million.
Fear as the Engine of Global Surveillance
The book argues that humanity has no right to make mistakes, and therefore "fixing" AI is useless — its creation must be stopped. The measures they propose go far beyond ordinary regulation. We are talking about global surveillance of every powerful processor, a ban on publishing research that reduces the cost of training AI, and, as a culmination, military strikes against violating countries.
The authors cite polls where 69% of Americans and 60% of Britons support strict AI regulation. However, their recipe turns fear into a tool for building a system of total control, where "good intentions" justify any means. As Pavel Durov aptly noted, such appeals "activate very ancient and deep parts of our brain," forcing people to abandon rationality in order to protect "children."
Crypto as the Achilles' Heel
For us, members of the crypto community, there is a troubling nuance in this book. The authors directly state that a superintelligence could finance its actions by hacking crypto wallets or manipulating memecoins. They cite the hacks of Mt.Gox and Bybit, as well as the story of Truth Terminal. According to their logic, this makes cryptocurrency one of the tools of a potential apocalypse, and therefore another target for the global surveillance they propose.
The book ends with a prayer that it is wrong, but its radical proposals are already on the table. As an expert, I see in this not so much a solution to the AI problem, but a project for a society based on fear, where for the hypothetical salvation of the future, we are ready to sacrifice the present. This reminds me of an old truth: the road to hell is paved with good intentions, and in this case, the asphalt for that road is mixed with fear of the superintelligence.