Doomers and Accelerationists Debate AI Extinction Risk

El Pais reports that Eliezer Yudkowsky and Nate Soares publish a book titled If Anyone Builds It, Everyone Dies arguing that sufficiently advanced artificial intelligence could produce human extinction, potentially "in a matter of months or within a decade," per the article. El Pais records that the authors were among signatories of a wider open letter in 2023 calling for a six-month moratorium on certain AI research. The article also cites a prior Yudkowsky writing that advocated severe enforcement measures, quoting that violators should have their data centers "destroyed by air strike," as reported by El Pais.

El Pais places the argument in the context of recent improvements in transformer-based systems such as ChatGPT, Gemini, and Sora, noting those advances have intensified public debate. Industry-pattern observations: discussions about catastrophic risk typically hinge on uncertainty about scaling behaviors, reward specification, and capability generalization as models grow larger and more autonomous. These are generic technical fault lines seen across AI-safety literature and reporting.

Editorial analysis: The El Pais article frames the debate as a polarization between "catastrophists" who emphasize existential risk and "accelerationists" who stress transformative benefits like curing disease and increasing productivity. For practitioners, this polarization drives competing policy proposals, including research moratoria, international coordination, and compute limits - themes visible in public discourse and regulatory proposals.

Editorial analysis: Observers should track formal proposals for international treaties or compute limits, influential op-eds and signatories from the research community, and any concrete regulatory text emerging from national governments. Also watch how mainstream technical conferences and major labs respond publicly to high-profile claims about extinction risk, and whether those responses change funding, red-teaming, or release practices.

El Pais presents the authors' claims and the contrasting accelerationist position. The article reports these positions and excerpts from prior writings; it does not provide new empirical evidence that resolves the underlying technical uncertainties about model capabilities or alignment.