[pre-Announcement] Bryan Caplan!
Sunday 15 June 10AM-1PM
Join us: Sunday 15th June
10AM - 1PM
in Justanotherspace, Nakameguro, Tokyo.
Renowned GMU economist Bryan Caplan, author of Build, Baby, Build and You Will Not Stampede Me: Essays on Non-Conformism sits down for a data‑driven conversation with the Tokyo rationality community on the challenges of betting on AI Outcomes and AI Risks, building on Bryan’s famous 2017 bet with Eliezer Yudkowsky (to be resolved if anyone survives to 2030).
Caplan’s fearless empiricism makes him a natural partner for the rationalist audiences. Expect hard data, contrarian takes, ideas so useful they might be illegal, and plenty of Q&A.
Bryan’ll be thinking in public in this, his Third Dialogue with Tokyo rationalists.
Come ready to challenge, refine, and extend the arguments.
Some thoughts on betting on AI:
A viable wager needs (i) an objectively checkable event, (ii) a horizon short enough to survive, (iii) counterparties who agree on pay-out conditions and (iv) infrastructure to enforce settlement. Catastrophic-risk narratives usually fail at all four hurdles, so the rational “bet” often collapses to advocacy, research or career positioning rather than a financial trade.
Further, turning “AI doom” into a tradeable position, specifically, suffers from: “the world ends” / “a box of Nobel laureates” or similar is not a clear, settled-outcome event to contract on; it’s also unfalsifiable if “nothing measurable happens until it is too late”; there are severe timing and path-dependency risk, as you have to survive margin calls while the market stays “irrational” by betting on the market’s own survival; on the opposite side, pay-out for the doom position is extremely irrelevant; similarly, there are counterparty and liquidity problems if you cannot persuade anyone then there’s no liquid market to clear your position; instrument engineering costs for bespoke betting structures;
A pessimist might conclude that betting is futile, but practical partial hedges on intermediate scenarios exist. You can:
• Use volatility overlays: Own VIX calls or structured equity collars; they pay out if uncertainty about AI jolts overall risk premia, regardless of whether the worst case arrives.
• Make sector-specific shorts: Short labor-intensive industries most exposed to rapid automation (for example, BPO and low-margin software services) while pairing with longs in AI-infrastructure names. This expresses the view that transition pain precedes either utopia or doom.
• Get catastrophe-linked reinsurance notes: Push insurers to include “AI-caused systemic outage” as a trigger, then buy the bonds. They may never exist, but lobbying for them is cheaper than bespoke derivatives.
• Explore reputation markets: Public, timestamped forecasts on platforms like Manifold or Metaculus create a non-cash call option on your credibility; if you are right, career and influence dividends may exceed any monetary hedge.
These instruments are imperfect, but they meet the minimum criterion of having a defined trigger and a solvent counterparty. They will not save you from literal extinction, yet they can align portfolios with intermediate scenarios—slower crashes, regulatory shocks, or mass realisation of unaligned AI risk—long before the lights go out.