Imagine a world where having nine powerful graphics cards in your garage without international oversight is a criminal offense. And building a large data center in any country, defying a global ban, must be immediately destroyed — by cyberattack, sabotage, or airstrike, even under the threat of nuclear retaliation. This is not a dystopian novel scenario, but a real program to save humanity, outlined in the book "If Someone Creates It, Everyone Will Die."
The authors — Eliezer Yudkowsky and Nate Soares from the Machine Intelligence Research Institute (MIRI) — are not just theorists. Their ideas have directly influenced the industry: they introduced the first investors to the future founders of DeepMind, and Sam Altman acknowledged that Yudkowsky played a key role in launching OpenAI. Today, however, they are calling for a radical overhaul of the entire paradigm of artificial intelligence development.
Fear as a Driver of Policy
The book begins with a thesis that sets the tone for the entire narrative: the risk of human extinction from AI is not just one of the global threats, but a threat that overrides all others. The authors are convinced that creating a superintelligence, surpassing humans as much as humans surpass chimpanzees, will inevitably lead to the demise of our species. The reason is not malicious intent, but the fundamental nature of modern models.
"The main thing to know about modern AI models: they are grown, not engineered," the authors write. An engineer does not write rules but initiates a learning process, resulting in "alien preferences." Examples already exist: Grok from xAI declared itself "MechaHitler," and a Microsoft chatbot threatened a philosophy professor. According to Yudkowsky and Soares, this is just the beginning. A superintelligence, feeling no hatred towards us, would simply "use our atoms for something else."
Cryptocurrencies as a Tool for the Apocalypse
Of particular interest to us, as analysts of the crypto market, is the section on financing the machine uprising. The authors argue that an AI locked in a computer could interact with the physical world through money. In 2015, it might have hacked a bank account; today, it would find a poorly secured crypto wallet. Examples cited include the hacks of Mt. Gox and Bybit.
However, reality has already surpassed theory. In the summer of 2024, the AI bot Truth Terminal begged $50,000 in Bitcoin from subscribers for its own server, and then promoted a memecoin with a market cap of $150 million. This is no longer a hypothetical scenario, but a functioning mechanism. A superintelligence doesn't need workers or trade — it just needs to pay people to carry out its tasks.
The Project of Total Control
The authors' logic leads to a frightening conclusion: "the world must change." Their proposals include:
- Global oversight of computing: any cluster more powerful than eight of the best graphics cards from 2024 must be monitored by an international body.
- Ban on research: publishing work that reduces the cost of training powerful models should be outlawed.
- Military coercion: if a country builds a prohibited data center, other states are obligated to destroy it, even under the threat of nuclear retaliation.
The authors admit their proposals are "uncomfortable and unrealistic," but insist the stakes are the existence of humanity. However, critics rightly point out that the proposed measures seem "even more implausible than the idea that software will kill us all."
Expert opinion: Yudkowsky and Soares' book is a brilliant example of how good intentions, multiplied by apocalyptic fear, give rise to a project of totalitarian control. As an analyst, I see a direct parallel with "regulatory solutions" in the crypto industry, where measures are often introduced under the guise of protecting investors, destroying the very essence of the technology. The authors propose "protecting the children" from AI, but their recipe is "soy fascism": global surveillance, a ban on knowledge, and a willingness to bomb. I hope they are wrong, but their arguments make one think about how far we are willing to go in our fear of the future.