New paper warns ASI by 2033: five existential risk challenges identified
AGI predicted as early as 2028, ASI by 2033 — are we prepared?
Matthew Baxter's new paper, hosted on Zenodo (DOI 10.5281/zenodo.20779325), lays out a compressed timeline for superintelligence: AGI by 2028–2030 and ASI by 2029–2033, driven by recursive self-improvement. While acknowledging these dates are controversial, Baxter uses them as a reference to gauge how much time remains to address the five core challenges he identifies. The paper argues that the field’s current focus on technical alignment alone is insufficient — each of the five dimensions must be solved in concert to avoid existential risk.
The five challenges are: (1) technical alignment and safety (the classic control problem), (2) concentration of power (who controls ASI and how it centralizes influence), (3) international governance (avoiding arms races and rogue actors), (4) social health (meaning, relationships, and economic disruption from automation), and (5) ethical challenges (value pluralism and moral frameworks). Baxter contends that a major failure in any one of these areas could cascade into existential catastrophe. The paper is a systematic attempt to map the full landscape of ASI risk beyond the narrow alignment debate, pushing the community to adopt a more holistic preparedness strategy.
- Timeline prediction: AGI by 2028-2030, ASI by 2029-2033 based on recursive self-improvement loops.
- Five identified challenges: technical alignment, power concentration, international governance, social health, ethics.
- Claim: Failure in any single challenge carries existential risk — technical safety alone is insufficient.
Why It Matters
Forces AI safety researchers and policymakers to broaden focus beyond alignment to systemic societal risks.