Dwarkesh PodcastFin Moorhouse - Longtermism, Space, & Entrepreneurship
CHAPTERS
- 0:00 – 0:54
Preview: anti-fragile belief-updating, earning to give, and many-worlds ethics
Fin previews several core themes: treating being wrong as a feature (not a bug), the logic of making money to later do the most good, and whether the many-worlds interpretation changes what matters. The teaser sets up the episode’s blend of EA strategy, decision theory, and big-picture futurism.
- •Anti-fragility: publicly endorsing mind-changing and correction
- •Earning large sums as a tool for later doing maximal good
- •Diminishing returns to personal wealth vs. (potentially) not for world-improvement
- •Many-worlds interpretation as a possible decision-relevant fact
- 0:54 – 3:39
Fin’s path into EA: philosophy, community pull, and pandemic timing
Dwarkesh introduces Fin’s roles at Oxford/FHI and Hear This Idea. Fin explains how philosophy (Singer/MacAskill) plus a strong local community drew him into EA, and how pandemic timing gave him space to skill-build and apply for opportunities.
- •Motivation via philosophical consistency and ethical reflection
- •Importance of finding an in-person EA student community
- •Pandemic as a ‘forced sabbatical’ enabling exploration and learning to code
- •Applying widely with low expected odds and getting lucky
- 3:39 – 10:41
EA writing & criticism prizes: incentivizing red-teaming before ideas ossify
They discuss prizes for EA-related writing and for critiques of EA. Fin frames criticism as essential for movement-level epistemics, especially during growth phases when beliefs can become socially entrenched and costly to question.
- •Why fund criticism: improve beliefs and prevent movement failure modes
- •‘Plasticity’ early on vs. later ossification of trendy beliefs
- •Celebrating shutting down projects as a success when evidence turns negative
- •Analogy to for-profit feedback loops vs. weaker nonprofit feedback
- 10:41 – 21:42
Longtermism’s late arrival: missing conceptual tools and new ways of thinking
Dwarkesh asks why “the future could be huge, so it matters a lot” took so long to formalize. Fin argues that key conceptual apparatus—probability, formal risk thinking, and the idea history could end—are surprisingly recent, making longtermism less obvious historically.
- •Tom Moynihan’s work on the intellectual history of existential risk
- •Probability and formal expected-value reasoning as shockingly modern tools
- •Novelty/precedent-free risks stand out as especially worth attention
- •‘Boundless progress’ framing: new ideas create new problems (Deutsch)
- 21:42 – 31:17
For-profit vs. nonprofit impact: markets, missing incentives, and ‘tails coming apart’
They explore whether EA underrates building profitable companies that also create social value. Fin emphasizes where nonprofit work dominates (future people, public goods, free-rider problems), while Dwarkesh argues markets are far from fully efficient given founder scarcity and context gaps.
- •Why nonprofits can be more counterfactually impactful when markets don’t form
- •Creating markets (AMCs, prizes) to import competitive feedback loops
- •Founder scarcity + neglected-context opportunities (e.g., Africa) can beat ‘efficient market’ intuition
- •Earning then optimizing separately vs. ‘social enterprise’ mixing objectives
- 31:17 – 36:48
Backtesting EA and decision strategy: EV vs minimax regret vs Kelly-style constraints
Dwarkesh asks how EA heuristics would have performed across history and whether they can misfire. Fin distinguishes ex ante rationality from ex post outcomes, arguing that high-EV ‘moonshots’ can look like failures in hindsight, and introduces minimax regret and Kelly-criterion analogies for bankroll/ruin constraints.
- •Correct strategies can still yield many ‘failures’ due to outcome variance
- •Minimax regret vs. maximizing expected value as different decision rules
- •Kelly criterion as a way to avoid ruin while still taking advantageous bets
- •Acting ‘on the margin’ (within a larger funding ecosystem) changes risk calculus
- 36:48 – 39:00
EA billionaires and risk attitudes: diminishing returns for selves vs. the world
Building on Dwarkesh’s “more EA billionaires” thesis, Fin argues altruistic motivation can rationally increase risk tolerance because the world doesn’t face the same diminishing returns as personal consumption. They discuss why EA demographics (young, techy, ambitious) could be overrepresented among future founders.
- •Risk aversion is rational for personal utility with steep diminishing returns
- •Impartial impact goals reduce diminishing-return concerns → more risk-neutral bets
- •Community clustering of ambitious people can increase startup formation
- •Big upside outcomes dominate when ‘doing good’ scale is enormous
- 39:00 – 51:37
Many-worlds interpretation: measure, identity, selection effects, and decision intuitions
Dwarkesh presses whether many-worlds should affect practical decisions, including existential risk intuitions. Fin offers a cautious ‘brain dump’: possible shifts toward risk-neutrality, altered personal identity intuitions, and anthropic-style puzzles (Sleeping Beauty variants) that resemble simulation-argument reasoning.
- •Many-worlds as ‘cloud of successors’ reframing personal identity
- •Measure-weighting vs. naïve branch-counting; probability language becomes tricky
- •Anthropic puzzles: 1 vs. 100 observer worlds and why 100/101 feels tempting
- •Implications for discounting/selection arguments and links to simulation reasoning
- 51:37 – 1:00:11
EA talent search and encouraging the young: wide nets, “bat signals,” and hubs
They discuss how EA can identify high-potential contributors globally and early. Fin references Tyler Cowen’s and Nick Whitaker’s ideas: instead of mass prestige filtering, create honest ‘bat signals’ that attract the right self-selected people, and invest in physical hubs/fellowships that compound motivation and mentorship.
- •Talent identification remains inefficient; big upside from better matching
- •Lamplight/bat-signal model vs. broad-prestige application funnels
- •Value of physical co-location: peers + senior proximity as motivation multiplier
- •Concerns about community-building adverse selection and movement presentation
- 1:00:11 – 1:21:58
The Long Reflection debate: coordination, avoiding irreversible lock-in, and governance risks
Dwarkesh challenges the long reflection as potentially requiring totalitarian enforcement and producing stagnation or captured governance. Fin reframes it as a directional ideal—more time before irreversible projects—focused on preventing unilateral lock-in rather than enforcing a single global plan, while acknowledging real risks of power entrenchment.
- •Motivation: current century as dangerously ‘unsustainably’ volatile
- •Irreversibility risk: space settlement/major projects could lock in values
- •Directional ideal: extend reflection time without needing a final decider
- •Tradeoff: consensus can reduce downside variance but may slow and be captured
- 1:21:58 – 1:38:51
Podcasting craft: why it works, why questions are hard, and underrated interview skills
Fin explains Hear This Idea’s origin as a low-cost way to talk to interesting academics, highlighting a ‘snowball’ effect in guest acquisition. They discuss why podcasts feel like natural conversation, why disfluencies can aid comprehension, and what makes questions good: specificity, cutting off meandering, and flushing out canned spiels.
- •Low startup cost + high academic ‘yes rate’ creates a compounding guest pipeline
- •Conversation as high-bandwidth idea transfer; transcripts can become text artifacts
- •Disfluencies and slight noise can increase comprehension via attention
- •Interviewing skills: specificity, timing, knowing what not to ask, and steering away from rehearsed answers
- 1:38:51 – 1:44:16
Space as an extinction hedge—and why risks aren’t independent
Dwarkesh pivots to space governance and whether space settlement makes extinction asymptote to zero. Fin argues diversification helps only if risks are independent; many key risks (pathogens with travel, unaligned AGI) correlate across locations, so space isn’t a sufficient hedge—though aligned AGI or improved coordination might reduce long-run peril.
- •Diversification logic: backups work only with independence assumptions
- •Pathogen spread and interplanetary travel undermine independence
- •AGI risk likely defeats simple ‘distance’ hedging
- •Possible end of ‘time of perils’ via aligned AGI or improved coordination/wisdom
- 1:44:16 – 1:47:28
Space conflict from first principles: attribution problems, first-strike incentives, and defense advantages
They explore whether space encourages war: Gwern’s argument that attribution is hard, weakening MAD and strengthening first-strike logic. Fin notes countervailing considerations such as detectability in open 3D space (‘nowhere to hide’) and points to Anders Sandberg’s longer-running work on these questions.
- •MAD relies on attribution + credible retaliation; space may break both
- •First-strike incentives increase if you can’t identify attackers reliably
- •Potential defensive advantages: visibility in space and difficulty of stealth
- •Cosmological isolation in the very long run (expansion) as a weak consolation
- 1:47:28 – 1:52:39
Von Neumann probes, ‘grabby aliens,’ and whether that implies a near-term space race
Dwarkesh asks if self-replicating probes inevitably burn the cosmic commons. Fin connects to Robin Hanson’s ‘Burning the Cosmic Commons’ and ‘Grabby Aliens,’ noting selection effects favor fast-expanding strategies but cautioning that these dynamics likely operate on timescales so vast that decade-scale racing is probably misguided.
- •Self-replication analogy to biological competition and containment dynamics
- •Selection effects may favor fast ‘grabby’ expansion strategies (Hanson)
- •Value question: avoid a future dominated by ‘dead’ resource-consuming expansion
- •Near-term implication skepticism: space is huge; marginal delays may be negligible
- 1:52:39 – 1:57:30
Practical space governance today: satellites, anti-satellite weapons, and orbital crowding
They return from far-future charters to nearer-term governance that could matter soon: regulating anti-satellite capabilities and managing rapidly increasing objects in orbit. Fin highlights data showing a ‘hockey stick’ in launches (especially LEO) and the institutional gaps around escalation and debris risks.
- •Long-run charters may not survive transformative AI; nearer-term issues still urgent
- •Anti-satellite weapons and weak regulatory consensus as a destabilizing risk
- •LEO crowding trends and the compounding governance burden
- •Value of better public datasets and measurement to guide policy
- 1:57:30 – 2:20:42
Startups, what EA may be missing, and career advice—then closing plugs for the prizes
Fin asks Dwarkesh what’s next; Dwarkesh leans toward building a startup unless the podcast scales dramatically. They discuss what EA underrates (translation layers for vague-but-important ideas; risks of dilution vs. planning for best-case growth), then offer practical career heuristics (avoid default paths, seek mentors, iterate until uncertainty stabilizes) before closing with details on the criticism contest and Effective Ideas prizes.
- •Dwarkesh’s 2024 plan: try a startup; remain podcasting if scale/impact jumps
- •EA ‘compiler’ problem: translating rich, ambiguous claims into tractable EA frames
- •Fin’s two worries: idea dilution/greenwashing vs. failing to plan for best-case scale
- •Career heuristics: don’t wait for certainty; seek mentorship; practice agency; use 80,000 Hours resources
