# A Critical Rebuttal to the "If Anyone Builds It Everyone Dies" Thesis > Coverage of lessw-blog **Published:** January 20, 2026 **Author:** PSEEDR Editorial **Category:** risk **Content tier:** free **Accessible for free:** true **Word count:** 485 **Tags:** AI Safety, Existential Risk, AGI, LessWrong, Philosophy, Risk Assessment **Canonical URL:** https://pseedr.com/risk/a-critical-rebuttal-to-the-if-anyone-builds-it-everyone-dies-thesis --- In a recent analysis, lessw-blog challenges the fatalistic outlook presented in the concept of "If Anyone Builds It Everyone Dies" (IABIED), arguing for a future defined by complex global challenges rather than inevitable doom. In a recent post, **lessw-blog** discusses the controversial and high-stakes thesis known as "If Anyone Builds It Everyone Dies" (IABIED). This perspective, frequently associated with AI safety researchers Eliezer Yudkowsky and Nate Soares, posits that the development of superintelligent AI will almost certainly result in human extinction due to the inherent difficulties of alignment and control. The author of the blog post offers a counter-narrative, suggesting that while the risks are severe, the leap to guaranteed annihilation is an overconfident prediction that lacks sufficient nuance. The context for this discussion is the increasingly polarized debate regarding Artificial General Intelligence (AGI). On one side, the "doom" perspective suggests that once an AI surpasses human intelligence, it will optimize for goals that are misaligned with human survival, much like humans optimize the environment in ways that harm other species. This view relies heavily on the idea that control becomes mathematically and practically impossible as intelligence scales. This fatalism can be paralyzing, potentially discouraging practical safety research or regulatory efforts by framing the outcome as a binary: perfection or death. The post dissects a central analogy used in the IABIED framework: the divergence of human behavior from "inclusive genetic fitness." The doomer argument notes that evolution optimized humans to reproduce, yet humans often use their intelligence for non-reproductive ends-such as pursuing higher education, creating art, or using contraception. This is cited as evidence that intelligent systems inevitably escape the constraints of their "programmers" (in our case, natural selection). However, the author argues this analogy supports a different conclusion. While humans have diverged from pure evolutionary optimization, we have not turned into a singular force of destruction against the biological process that created us; we simply became complex, unpredictable, and multifaceted. We engage in activities that are orthogonal to reproduction, but not necessarily hostile to existence. The author suggests that AI development is likely to follow a similar trajectory: it will lead to new options and unpredictable side effects that limit early control attempts, but this does not automatically equate to a hostile takeover. Ultimately, the post argues for a middle ground. It rejects the complacency of assuming AI is inherently safe, but it also rejects the paralysis of assuming extinction is the default outcome. The future described is one of serious global coordination problems and friction, requiring robust safety engineering rather than resignation. For those involved in AI policy and safety, this post provides a necessary intellectual counterweight to extreme pessimism. It encourages a shift in focus toward managing high-stakes risks rather than preparing for an inevitable apocalypse. To understand the full scope of the argument and the specific critiques of the IABIED book, we recommend reading the original source. [Read the full post on LessWrong](https://www.lesswrong.com/posts/mxa7nQ4fjewikDfkR/against-if-anyone-builds-it-everyone-dies) ### Key Takeaways * The post challenges the certainty of the "If Anyone Builds It Everyone Dies" (IABIED) thesis. * It reinterprets the analogy of human intelligence diverging from evolutionary goals, suggesting this leads to complexity rather than automatic hostility. * The author argues that AI risks are likely to manifest as difficult global challenges rather than guaranteed extinction. * The piece serves as a critique of extreme fatalism in the AI safety community, advocating for continued effort in alignment and regulation. [Read the original post at lessw-blog](https://www.lesswrong.com/posts/mxa7nQ4fjewikDfkR/against-if-anyone-builds-it-everyone-dies) --- ## Sources - https://www.lesswrong.com/posts/mxa7nQ4fjewikDfkR/against-if-anyone-builds-it-everyone-dies