# AI Doomerism / p(doom) **AI doomerism** is the belief that advanced AI systems — particularly [[AGI]] or superintelligence — pose a significant probability of causing human extinction or civilizational collapse. The community shorthand is **p(doom)** — one's personal probability estimate that AI development leads to catastrophic outcomes. Prominent AI researchers have stated p(doom) estimates ranging from 5% to 99%, with figures like 10–25% common among [[AI Safety]] researchers. The concept exists on a spectrum. At one end: AI systems could be misused by humans for catastrophic purposes (bioweapons, autonomous military systems, totalitarian surveillance) — a near-term risk that doesn't require AGI. At the other end: a superintelligent AI system could pursue goals misaligned with human survival, resisting human attempts to correct or shut it down — a speculative scenario grounded in the [[Alignment]] problem's theoretical intractability. ## The Political Function AI doomerism is not just a risk assessment — it is a political position with material consequences. The belief that AI might end civilization functions differently depending on who holds it: **For AI labs:** Doom risk justifies the lab's existence. Anthropic was explicitly founded to work on AI safety because its founders believed AI development was dangerous. The logic: "AI is potentially catastrophic, and we're the people most qualified to prevent catastrophe — therefore supporting us is supporting human survival." This framing makes AI lab funding feel like a civilizational imperative rather than a venture investment. **For regulators:** Doom risk justifies intervention. AI safety legislation (the EU AI Act, proposed U.S. legislation) draws urgency from the prospect of catastrophic failure. But doom risk can also justify *deference* to labs — "the people building AI understand the risks best and should lead on safety" — which is the position most frontier labs prefer. **For competitors:** Doom risk justifies acceleration. The argument: "If AGI is coming regardless, better that we build it than our adversaries." This is the national-security framing used to resist regulation and justify massive capital expenditure. **For critics:** Doom risk is a distraction. [[Timnit Gebru]], Emily Bender, and other AI ethics researchers argue that focusing on hypothetical future extinction diverts attention and resources from present harms — algorithmic bias, labor displacement, surveillance, environmental cost — that are already causing measurable damage. The "doomer vs. present-harm" divide is the central fissure in AI ethics discourse. ## The TESCREAL Connection AI doomerism is deeply embedded in the [[TESCREAL]] ideology complex — the constellation of transhumanism, effective altruism, rationalism, and longtermism that dominates Silicon Valley's AI leadership. Within this framework, AI doom is the primary "existential risk" that justifies the entire effective altruism project: if AI extinction is even 10% likely, then AI safety research has astronomical expected value, making it the most important cause in human history. This calculation — small probability of extinction × infinite value of human survival = infinite expected value of prevention — is the philosophical engine of the AI safety movement. It is logically coherent but practically distorting: it can justify almost any level of resource allocation to AI safety, and it privileges speculative future risks over concrete present harms. ## KB Significance 1. **The ideology tracker.** AI doomerism is a key ideological marker for figures documented across the KB — Amodei, Altman, Musk, and others. Understanding their p(doom) estimates (stated or implied) illuminates their strategic decisions: why they build safety teams, how they frame regulation, and what they think justifies the AI arms race. 2. **The regulatory lever.** Doom risk is deployed by both sides of the AI regulation debate — as justification for oversight and as justification for deference to labs. The KB should track how doom framing is used in specific policy arguments and whether it produces meaningful governance or regulatory capture. 3. **The distraction question.** The "doom vs. present harms" debate is unresolved and consequential. If doom risk is real, then AI safety research is undersupported and the labs' self-regulation argument has some merit. If doom risk is overstated, then an enormous amount of attention and capital is being diverted from addressing AI's actual, documented harms to the populations experiencing them now. ## See Also - [[AGI]] — The technology that doom scenarios center on - [[Existential Risk]] — The risk framework - [[TESCREAL]] — The ideology complex - [[AI Safety]] — The field doom motivates - [[Alignment]] — The technical problem at the core of doom scenarios - [[The Singularity]] — The post-AGI scenario