Doomsday tech: could AI really kill us all? - France 24

TEXT START: Tokyo (AFP) – Stark warnings from artificial intelligence researchers that advanced forms of the technology could wipe out humanity have raised questions over whether governments are moving fast enough to regulate it.

The Dissection

The article packages AI risk as an extinction debate: rogue superintelligence, engineered pathogens, nuclear escalation, hacking, and atmospheric destruction. It correctly exposes the competitive logic driving laboratories forward despite acknowledged danger. But it gives the most dramatic possibility—human extinction—the central stage while treating job losses and economic displacement as secondary problems.

The text is really performing risk domestication. It makes AI appear dangerous enough to require licensing, yet inevitable enough that states and firms must keep racing. The result is a spectacle of imminent catastrophe that leaves ownership, control of AI capital, and the collapse of human bargaining power largely untouched.

The Core Fallacy

The article assumes that AI becomes systemically terminal only if a superintelligence physically kills humanity. That is the wrong threshold.

Under the Discontinuity Thesis, the decisive sequence is cognitive automation dominance, coordination failure, and productive participation collapse. Once AI performs economically necessary cognitive work at lower cost and superior speed, the mass employment-to-wage-to-consumption circuit breaks. Humans can remain alive while becoming economically unnecessary. That is already a terminal event for the post-WWII order.

The article’s focus on extinction therefore creates a false hierarchy: speculative human annihilation is treated as the ultimate danger, while the more structurally immediate destruction of broad human economic participation is demoted to a conventional policy concern. Regulation may delay the transition. It cannot restore the old circuit once competitive AI ownership makes human labor redundant.

Hidden Assumptions

• Superintelligence is required for AI to become civilization-threatening.
• Governments can coordinate faster than firms and states compete.
• Licensing can contain capabilities without merely shifting them to less constrained rivals.
• Preserving human survival preserves the existing economic order.
• Job losses are a side effect rather than the central mechanism of systemic death.
• Firms can voluntarily slow development despite a race for strategic and financial dominance.
• Human oversight remains meaningful as systems become more capable.
• Regulation can manage ownership and distribution without confronting concentrated control of AI capital.
• Extinction-risk percentages provide more useful guidance than structural analysis of labor displacement and power concentration.
• Consumption can be maintained through transfers without addressing the loss of productive participation.

Social Function

Classification: partial truth, transition management, prestige signaling, and ideological anesthetic.

The article tells a real story about competitive escalation, misuse, and weak coordination. Its distortion lies in making the end of humanity the benchmark for seriousness. That framing lets elites debate whether machines might kill everyone while avoiding the less cinematic question of who will own the machines after most people are no longer economically necessary.

It also provides reputational cover. Researchers can present themselves as responsible guardians of civilization, companies can use danger warnings to signal technological importance, and governments can discuss agile regulation without resolving the underlying ownership conflict. The apocalypse becomes a bargaining instrument; the labor market becomes collateral damage.

The Verdict

The article identifies a genuine race dynamic but misdiagnoses the terminal mechanism. AI does not need to exterminate humanity to kill the postwar economic order. It only needs to sever the requirement that most humans sell labor to survive.

The extinction scenario is the headline-sized corpse. The deeper death is productive participation collapse: humans remain alive, consume if transfers are politically sustained, and lose their status as necessary economic agents. The article sees the fire in the laboratory but mistakes the burning building for the only threat.