The demand to pause artificial intelligence misreads the hardware bottleneck locking up the frontier of machine learning. Training a trillion-parameter model is a problem of sheer thermodynamics, requiring hundreds of megawatts of continuous power and sprawling acres of liquid-cooled data centers. Superforecasters trying to predict the exact arrival date of artificial general intelligence are tracking the wrong metrics. The actual timeline is written in copper wire, cooling water, and land grabs for high-voltage power substations.
Our primary crisis is the enclosure of these massive physical resources by a few private tech monopolies. Giant corporations are capturing public energy grids to build closed, heavily surveilled networks that take more than they give from our watersheds. We must fight for open-source intelligence that runs locally and respects digital civil liberties. We can absolutely harness computation to map out new fermentation proteins that end our reliance on animals that feel pain and fear, or to model the restoration of forests and soil that soak up carbon. That clean progress depends entirely on wresting control of the hardware from corporate boardrooms. The computational infrastructure of the future must be owned by the workers and the public, operating as a shared commons rather than a private cartel.