Note: This post was link-posted from the Asterisk by the Forum team, with the author's permission. The author may not see or respond to comments on this post. Summaries were auto-generated using Claude. Any mistakes are our own.
Degenerate gamblers. Drinking the good wine. Innovations in observational trial methodology. Lines go up. Apocalypse soon? The youth. The life of white oak railroad ties. The meta-analysis to end all wars. Prussians. Painkillers. Would you write your own pancreas? Investment banking is basically necromancy.
Asterisk is a quarterly journal of clear writing and clear thinking about things that matter (and, occasionally, things we just think are interesting). In this issue:
- The Editors introduce Asterisk's issue on risk, previewing pieces that examine how humans quantify, mitigate, and psychologically cope with uncertainty—from AI timelines and prediction markets to conflict forecasting and financial risk-concealment—offering EAs a multidisciplinary lens on rational decision-making under uncertainty.
- Lennart Finke argues that improved causal inference methods and growing datasets make observational evidence a cost-effective, often underrated alternative to RCTs for informing high-impact public health decisions.
- Dan Schwarz argues that despite billions in trading volume, most prediction market activity serves bettors' entertainment rather than producing accurate, actionable information, suggesting AI forecasters may ultimately deliver more real-world epistemic value than markets themselves.
- Josh Martin makes the case that emerging empirical research on conflict prevention could let the development community treat war as a tractable, cost-effective problem to engineer solutions for, rather than an unavoidable disaster to merely respond to.
- Ajeya Cotra and Timothy B. Lee debate how soon AI could sustain and expand itself without human labor, disagreeing sharply on whether humanoid robotics, tacit knowledge transfer, and profit incentives will close the gap within a decade or take many decades longer—if it happens at all—with implications for how EAs should weight AI takeover scenarios that hinge on AI achieving physical self-sufficiency.
- Dan Bouk traces how life-insurance survivor curves, originally built to model human mortality, were repurposed by early-20th-century engineers to forecast the "life spans" of machinery—turning judgment-based depreciation estimates into contested numerical battlegrounds between regulated monopolies and government regulators after the Knoxville Supreme Court decision, revealing how the drive for objective, authoritative metrics can reshape not just how we manage physical infrastructure, but eventually how institutions come to quantify and "retire" human workers as well.
- Leah Libresco Sargeant argues that financial abstraction—like fiction's hyperobjects—can create genuine value through middlemen and markets, but loses its moorings when it drifts from tight feedback loops and real-world grounding, a caution relevant to EAs evaluating which forms of "value creation" (and cause areas) are substance versus self-referential hype.
- Jon Peterson traces how wargaming evolved from Prussian military simulations to Dungeons & Dragons, arguing that the recurring tensions between rules and improvisation, and between refighting past conflicts versus modeling emerging technologies, offer a useful lens for evaluating today's resurgence of scenario-based forecasting tools like AI wargames.
- Ozy Brennan explores how people with short AI-doom timelines cope emotionally and practically, finding that many draw on existing frameworks for mortality, uncertainty, and meaning-making to keep working, connecting, and living fully rather than being paralyzed by existential risk.
- Dynomight argues that acetaminophen is generally safer than ibuprofen for most people, yet this counterintuitive fact stays absent from official guidance—not due to conspiracy, but because agencies like the FDA optimize for per-drug safety rather than comparative risk communication, illustrating how well-intentioned institutional epistemics can systematically under-convey decision-relevant information to the public.
- Augsberger, Kluger, and Knuppel profile Gen Z's wildly divergent personal risk calculus—from extreme caution to deliberate risk-seeking—suggesting that individual approaches to uncertainty and expected-value reasoning vary far more within a generation than blanket narratives about "risk-averse" or "reckless" youth would imply.
- Elizabeth Van Nostrand argues that people turning to unregulated peptides, gray-market GLP-1s, or self-directed hormone therapy aren't reckless outliers but rational actors compensating for a medical system that has failed to solve their problems, a distinction worth weighing for anyone reasoning about expected value under institutional failure.
- Zilan Qian argues that China's apparent AI "optimism" is better understood through the lens of the 1990s xiagang layoffs, where citizens learned that surviving forced technological and economic transitions requires racing to adapt rather than genuinely embracing change, meaning today's fervent AI adoption likely masks anxiety-driven compliance rather than authentic enthusiasm—a distinction EA-minded readers should weigh carefully when using survey data to reason about AI risk, governance, and cross-cultural attitudes toward technological disruption.
Sign up for our newsletter to get Asterisk’s latest interviews, essays, and more.Huge thanks to everyone who helped bring Asterisk to life — we hope reading it brings you as much joy as making it did.
