Disaster is still better than AI

The essay argues that an AI-caused financial or infrastructure crisis, though painful, might be preferable to AI continuing to expand without an obvious catastrophe. In the author’s view, gradual adoption would make people dependent on AI while imposing long-term costs: heavy energy use, damage to the natural world, and the weakening of intellectual communities. People, he argues, respond to immediate harm more readily than distant risks, so efforts to make AI safer or more ethical may instead reassure the public and allow the technology to become harder to abandon. He rejects the idea that safeguards can make automated creativity ethical in itself, comparing genuine learning about danger to the instinct to avoid a hot surface after being burned.

The argument extends to academic mathematics, which the author says was already too focused on producing theorems rather than using reasoning to benefit people. AI’s disruption of the field could therefore create an opening to rebuild it around more human purposes. More broadly, he believes people could recover from a financial breakdown or supply-chain shock and create better systems, while media, technology executives, and wealthy elites have incentives to exaggerate the risks of disruption. A crisis might prompt the public to shut down data centers and technology companies; in the author’s view, the resulting world could be better than one increasingly organized around AI.