What if the singularity lies beyond a plateau we cannot cross? (www.jasonwillems.com)

A thought-provoking essay argues that instead of an imminent “AI singularity” driven by runaway recursive self-improvement, humanity may be approaching a broad, hard-to-cross plateau where progress slows dramatically. The author contends that many remaining bottlenecks are macro-scale—energy grids, semiconductor fabs, data centers and other infrastructure—that require massive capital, regulation, and logistics rather than small-team ingenuity. If true, the expectation of perpetual exponential gains in compute, energy efficiency, or space travel is optimistic; the singularity might lie beyond a set of physical, economic, and regulatory ceilings we cannot easily breach. Technically, the piece points to S-curve dynamics and hard physical limits (speed of light, thermodynamics, quantum tunneling as transistors approach atomic scales) that produce diminishing returns across domains such as cooling, solar efficiency, and transistor density. Breakthroughs (fusion, useful quantum computing, room‑temperature superconductors) remain possible but uncertain in timing and scope. Economic gravity—rising marginal costs per advance and the growing role of government-scale projects—further constrains progress. The practical implication for AI/ML is a recalibration of risk and strategy: prioritize near-term, high-impact gains that fit within physical and economic limits, and temper expectations for transformative, fast-moving change without clear breakthroughs.