Effective Altruism

Definition

Effective altruism is an impact-oriented framework that compares uses of money, time, and careers by the amount of good they are expected to produce.

Current Synthesis

Across the bounded sources, effective altruism begins with measurable present-day interventions, extends into career choice and neglected causes, and then expands through longtermism toward pandemics, nuclear war, uncontrolled AI, and other future catastrophes. Its language of expected impact, agency, and probability-weighted choice can motivate direct giving, useful skill-building, research, resignation, insider work, or institution-building.

The new source makes this evolution institutional rather than merely intellectual. Funding, fellowships, jobs, venues, and networks helped AI safety become a Silicon Valley and policy agenda. That success changes allocation: a cause can become less neglected, institutions can change direction, and methods developed for measurable health interventions may fail in strategic contests among companies, states, and political actors.

The framework does not determine one AI policy or excuse the means used in pursuit of impact. Sources range from centralized frontier control to decentralized defense, and FTX shows how future benefit can rationalize present misconduct. Durable effective altruism needs legitimate means, beneficiary-centered evidence, dynamic neglectedness assessment, institutional accountability, and humility about disputed probabilities.

Key Claims

• Resources and careers should be compared by expected positive impact rather than intention alone.
• Cause priority depends on importance, tractability, and changing neglectedness.
• The movement expanded from direct aid toward career strategy, longtermism, existential risk, and AI safety.
• Funding and social infrastructure can turn an intellectual priority into a research and policy field.
• Expected-impact reasoning can support opposed tactics because assumptions remain disputed.
• Impact maximization can motivate action but become overconfident, demoralizing, or a moral rationalization.
• Expected good does not excuse harmful means, and donor intention does not substitute for beneficiary outcomes.

Evidence

• AI-control critique: Anthropic’s $2T IPO, Zuck’s AI Manifesto, Nvidia’s $500B AI Bet, Grok’s Comeback contrasts an Anthropic-linked EA posture with Decentralized AI Control.
• Career application: Brave New whirl: Turkey’s opposition overhaul uses Benjamin Todd and 80,000 Hours to frame useful skill-building and contribution.
• Worker culture: Inside the mindset of AI safety workers connects AI-safety careers to expected impact, agency, and numerical uncertainty.
• Scandal: Dunes and gloom: MBS’s annus horribilis pairs Sam Bankman-Fried’s partial reassessment and denial with post-FTX movement persistence.
• Direct giving: E253 AI 会改变世界，可它真的有让世界变好吗？ applies effectiveness comparison to present interventions and delivery losses.
• Development: E257｜对话Vitalik：AI安全背后的哲学思想，有效利他主义如何影响硅谷？ distinguishes EA from rationalism and uses Vitalik Buterin’s giving to show dynamic neglectedness and institutional drift.

Counterevidence & Qualifications

The sources do not provide a representative movement history, membership survey, or comparative audit. They mix political critique, career advice, fieldwork, a convicted founder’s self-interested account, and practitioner interviews. Precise probabilities can create false confidence; long horizons can obscure accountability; and health-intervention methods may not survive political conflict. Moral hazard and influence do not show all adherents share one motive or policy.

What Changed

• Added the staged development from measurable charity through longtermism, post-FTX crisis, and AI safety.
• Added funding, fellowships, jobs, venues, and networks as agenda-forming mechanisms.
• Made neglectedness dynamic as fields gain attention and institutions change.
• Strengthened the boundary between measurable intervention comparison and strategic conflict.
• Added d/acc as evidence that impact premises do not uniquely imply centralized AI control.

Related Concepts

• Philanthropic Neglectedness - dynamic marginal allocation criterion.
• Impact Career Choice - applies comparison to careers.
• Rationalist Community - overlapping epistemic culture with an independent origin.
• Existential Risk Probability Estimation - quantifies uncertainty for decisions.
• AI Safety Worker Culture - community reinforcing agency and effectiveness norms.
• Defensive Decentralized Acceleration - protective and distributed risk response.
• Decentralized AI Control - competing response to centralized AI control.