Underground Intelligence

Summary#

Iris Long's scenario concerns the "fourth territory," the subterranean world of accelerators, neutrino detectors, and waste repositories, in a future where only China and the United States can afford to build there. Europe's Future Circular Collider has been cancelled, China's CEPC is the new Higgs factory, and the deep-learning systems processing collider data have begun proposing particle models of their own. The frontier of technological geopolitics, Long writes, is "the clash of methods to interpret the universe."

The argument#

Long's first claim is infrastructural: fundamental physics now depends on computational surplus. The High-Luminosity LHC produces ten times its design luminosity and exceeds Europe's available compute, so even a European consensus to build the next machine could not run it, while China's overbuilt AI infrastructure from the last cycle powers its underground experiments. Her second claim is epistemic. Silicon Valley's newest models, which build predictive spaces from minimal data, ignore "backend" fields with long return cycles, so physics still runs older big-data algorithms. Those algorithms, trained on ultra-high-dimensional detector data that bears "no resemblance to anything human," have nonetheless begun to generate their own theories faster than humans can propose them. Machines built to verify theories now produce them. Her third claim follows: since particles cannot be restricted in flow and the two powers are roughly equal in algorithms, compute, and energy, the competition is over application, and each new model is a wager on energy, economic, or defence capacity.

Section by section#

The piece reads as an essay with a reportorial future tense and no characters. It moves from the FCC's suspension to the CEPC's placement on China's quantum-fibre science network, then to the machine intelligence itself, which "has no concept of 'human.'" The closing provocation is that many of the new models appear explainable, and in more than one way, so a theory of everything "might not exist. Perhaps it never has."

Key concepts#

• Fourth territory — the subterranean realm of underground laboratories, treated as a sovereign domain alongside land, sea, and orbit.
• Computational crisis — the condition in which an experiment's data output outruns a region's compute, making physics a function of AI infrastructure.
• Reversal in fundamental science — apparatus built to verify human theories instead generating theories of its own.

Connections#

Part of Hemispherical Stacks. Its non-human science intelligence, trained on data unlike anything human and arriving at models humans cannot match, belongs to the program's wider interest in intelligence that does not resemble us, argued directly in How to Think Unlike Humans. Xu's Martian Rovers in the same collection is its mirror: there scarcity of data forces a different intelligence, here abundance does.