There’s a physical version of the AI singularity, and it matters a lot

Robotics researcher Jeff Schneider argues that beside the feared software “AI singularity” of self-improving intelligence sits a nearer physical singularity: industrial capacity that recursively expands with less human labour. Machines already handle warehouse, mining and construction work; the threshold is integrating design, logistics, manufacturing and deployment so machines extract, transport and build factories that build more machines. In the next decade or two, labour could cease to be the main constraint on industrial growth—energy, materials and capital remain binding, but removing labour bottlenecks lets the system hit and mitigate them faster. Only “enough” AI to enable machine reproduction of industrial capacity is required, not infinite intelligence; control of design and deployment would confer compounding economic and material power. Schneider stresses physical AI tolerates less error than chatbots (a 90%-reliable robotaxi is dangerous) and urges policymakers to watch robotics, AVs, mining, energy, manufacturing and construction—not only models, chips and data centres—because early advantages could become irreversible, including in defence.