The C-Minus Warning
We barely survived industrialisation — AI-era equivalents are unsurvivable
The standard techno-optimist rebuttal to AI concern is that we survived industrialisation's disruptions and will adapt again. Harari's counter is that this claim is historically illiterate: adapting to industrialisation required European imperialism, Soviet communism, and Nazism. The 20th century's two world wars and multiple genocides were not accidents — they were the political experimentation the industrial revolution necessitated. Humanity passed, but barely, and with enormous cost.
The C-Minus Warning holds that the same class of experimentation will accompany AI's disruption — new surveillance empires, AI-enabled totalitarian regimes, and geopolitical ruptures driven by rapid labour displacement. The critical difference is that the 'failed experiments' of the industrial era were catastrophic but survivable. The equivalent experiments in the AI era — with nuclear-armed states, engineered pathogens, and AI-enabled autonomous weapons — are not survivable at civilisational scale.
The framework is not predictive of specific events but calibrates the realistic lower bound: 'It happened before, and we barely survived. If it happens again with more lethal tools, we don't.' This is a different risk framing than speculative AGI doom — it requires only historical pattern repetition, not science fiction.
- Past technological adaptation is not evidence of safe adaptation — the mechanism included catastrophic political experimentation that consumed hundreds of millions of lives
- The 'failed experiments' of each transition era scale with the lethality of available tools; industrial-era failures were survivable, AI-era equivalents may not be
- Speed asymmetry is the primary risk multiplier: industrial transitions gave decades for multiple warning shots; AI's timeline compresses this dramatically
- The return of war and rising defense budgets are the early observable symptoms of the same cycle, not anomalies
- Governance structures built for one technological era cannot govern the next — they must be rebuilt in parallel with, not after, deployment
- Reconstruct the full cost of prior transitionsBefore accepting 'we adapted before' as reassurance, map the complete cost of that adaptation: imperialism, totalitarianism, world wars, pandemics, famines driven by political experimentation. The optimist framing selects the endpoint, not the path.
- Identify the current cycle's early symptomsRising defense budgets, erosion of multilateral institutions, and the return of great-power competition are the observable early-cycle indicators that the same pattern is repeating. These are measurable in real time, not speculative.Pro tipDefense budget growth rates across NATO and non-NATO states are publicly reported quarterly. Sustained above-GDP growth rates across multiple states simultaneously is the empirically observable correlate of the transition Harari describes.
- Map the lethality upgrade between transitionsFor each class of 'failed experiment' from the industrial era, identify the AI-era equivalent and its lethality upgrade. Totalitarian surveillance states plus AI social scoring versus 20th-century equivalents. Autonomous weapons versus industrial-era military technology. This is the source of the asymmetric survivability claim.WarningAvoid mapping the upgrade as purely a weapons problem — AI-enabled financial control and information suppression have no industrial-era equivalent and carry their own lethality at civilisational scale.
- Assess the speed of governance constructionIndustrial-era governance (labour law, antitrust, central banking, international law) was built reactively over 50–100 years after the disruptions were visible. Evaluate whether current AI governance construction is happening in parallel or reactively, and estimate the lag.WarningIf governance is being built reactively, the window between disruption and stabilisation is where the dangerous experiments occur. That window is compressing.
- Calibrate your own risk tolerance against the historical baselineUse the C-Minus baseline — not utopia, not extinction, but the historical middle ground of catastrophic-but-survivable — as your realistic lower bound when assessing AI transition scenarios. Build positioning around the probability distribution between the historical lower bound and better outcomes, not between utopia and doom.
Harari uses the specific political experiments that industrialisation necessitated — European imperialism (colonies as resource capture and market expansion), Soviet communism ('how do you build an industrial society? You build a communist dictatorship'), and Nazism (totalitarianism enabled by industrial infrastructure including trains, electricity, and radio) — to demonstrate that 'successful adaptation' is not the right description. These were the adaptation mechanism.
Harari identifies 2023 as the point at which the most peaceful era in early 21st-century history ended, with global defense budgets rising rapidly across states that had been reducing military spending for decades. The vicious cycle: one nation arms, neighbours perceive threat and arm more, the first nation arms further. Money shifts from nurses and schools to tanks and missiles.
Harari developed this framework across Homo Deus (2015) and 21 Lessons for the 21st Century (2018), drawing on his professional expertise as a military historian and historian of the medieval and early modern periods. The industrial revolution serves as his primary empirical case — a technological transition for which we have complete historical data on the political experiments it generated.
In the DOAC interview, Harari uses it to explain why the 'we adapted before' argument is not reassuring but alarming: the full cost of prior adaptation is systematically excluded from the comparison. Pointing to the 2030s–2050s as the critical transition window based on his 2016 predictions being surpassed by 2023.