Technocalvinism (blog.cosmos-institute.org)

Luke Drago’s guest essay coins “technocalvinism” to push back against Mechanize’s fatalistic blog claim that “the future of AI is already written.” Mechanize argues that physical constraints and economic incentives make autonomous, job‑replacing agents inevitable and unavoidable — pointing to parallel discoveries like LLMs (compute scaling inferred from NVIDIA revenue and GPT‑4’s 2022 training) as evidence that tech follows a predetermined tree. Drago rebuts this by reframing technology as invented, not merely discovered: people and institutions make prioritization choices that shape what gets built. He lists four counterclaims — technology is invented, we can steer it, alternatives to full automation are possible, and full automation carries existential stakes — and uses historical counterexamples (absent wheels in Mesoamerica, lost Roman concrete, the Antikythera-like gaps) to show convergence isn’t a law. For the AI/ML community the piece matters because it challenges moral and strategic complacency: if development is steerable, engineers, funders and policymakers can prioritize defensive, human‑centric or “renewable” AI alternatives instead of accelerating full automation by default. Technically, the argument highlights distinctions between diffusion and invention, the role of compute and incentives in LLM timelines, and the feasibility of competitive architectures and governance choices that could change the order or nature of breakthroughs. The takeaway: treat AI trajectories as policy- and design-contestable rather than inevitable.