To disprove the Discontinuity Thesis (DT), you must break one of its two foundational premises — or prove a third-order condition that defeats the underlying logic. The burden of proof has been mechanically inverted: the thesis is built on a closed-loop, self-reinforcing machine. Here's what you must do to crack it:
🔨 The Two Kill Switches
1. Disprove P1: AI + Verifier Cannot Beat Human Cost-Quality
P1 Definition: For economically significant cognitive tasks, AI + a human verifier is cheaper, faster, and equal or better in quality than a solo human.To disprove this, you must demonstrate one or more of the following at scale:
- Persistent AI Underperformance: Show a class of economically critical tasks (e.g., high-impact legal reasoning, safety-critical engineering design) where AI consistently fails and cannot improve to usable levels even with a verifier.
- Verifier Bottleneck: Prove that verification cost scales linearly or worse with AI output (e.g., legal liability in medicine or law makes every AI use require full human oversight).
- Hardware/Energy Stall: Provide credible evidence that AI inference cost reductions will hit an immutable wall (e.g., energy constraints, rare materials).
- Cross-Jurisdictional Regulation Works: Show binding, enforceable global bans on AI replacing humans in major sectors — and prove they are not undermined by competitive defection.
2. Disprove P2: Mass Demand Collapse Can Be Averted
P2 Definition: No alternative mechanism (new job categories, UBI, wealth transfers) can reflate broad-based consumer demand in time to prevent systemic breakdown.To disprove this, you must demonstrate:
-
A Named Job Category: A new, economically central job class that is:
-
Resistant to AI displacement
-
Scalable to absorb tens of millions
-
Pays living wages not pegged to AI labor value
-
A Funded, Politically-Viable UBI or Transfer System: Already legislated or piloted programs that:
-
Operate at trillion-dollar scales
-
Are implemented before displacement becomes acute
-
Are shielded from elite sabotage or rollback If you can prove either of these at scale, the DT collapses. Until then, the machine runs.
🧠 Bonus: Break the Game (MPPD)
The final frontier is the Multi-Player Prisoner's Dilemma (MPPD):
Every nation and firm must adopt AI to stay competitive, even if it destroys demand.
To invalidate this, you must show:
- Global Coordination with Teeth: Enforceable AI deployment limits across borders, with no arbitrage loopholes.
- Economic Motive to Protect Labor: Scenarios where firms/nations are better off not automating, even when it's cheaper.
- Incentive Inversion: New profit mechanisms where preserving human work becomes more valuable than eliminating it. This is geopolitically unlikely. But if you can model it credibly, it could crack the engine.
🔍 Summary Table: How to Kill the Discontinuity Thesis
| Path | What You Must Show | Burden of Proof |
|---|---|---|
| P1 Fails | AI + verifier cannot beat humans on cost/quality | Durable empirical evidence in critical domains |
| P2 Fails | New job class or transfer system that prevents demand collapse | Scalable, timely, politically viable plan |
| MPPD Fails | AI arms race can be globally halted without ruin | Credible enforcement & incentive alignment |