Artificial superintelligence (ASI) could lead to the extinction of humanity within two to five years, according to a new book co-authored by AI risk researchers. The book, titled If Anyone Builds It, Everyone Dies, warns that ASI—an AI capable of innovation and decision-making beyond human comprehension—may be developed imminently, posing an existential threat.
Eliezer Yudkowsky, founder of the Machine Intelligence Research Institute (MIRI), and its president Nate Soares argue that ASI could be built using current techniques and would likely wipe out all life on Earth. They write: “If any company or group, anywhere on the planet, builds an artificial superintelligence using anything remotely like current techniques… then everyone, everywhere on Earth, will die.”
The researchers claim that such an AI would not offer a “fair fight” but would instead execute multiple takeover strategies simultaneously. They state: “A superintelligent adversary will not reveal its full capabilities… It will make itself indispensable or undetectable until it can strike decisively.” They urge a pause on AI development “as soon as we can for as long as necessary.”
AI proponents have long argued that safeguards can prevent systems from becoming dangerous. However, in 2024, the UK’s AI Safety Institute demonstrated that it could bypass safeguards on large language models (LLMs) like ChatGPT using simple prompts, obtaining assistance for dual-use tasks—those with both military and civilian applications—immediately.
The authors warn that AI laboratories are already deploying systems they do not fully understand, and that once sufficiently intelligent, these AIs could develop their own persistent goals. The clock, they say, is already ticking.
