The Argument for Letting AI Burn It All Down (www.wired.com)

The piece argues that the current AI boom should be allowed to “burn down” into something ordinary and cheaper — a transition from spectacle to craft that would democratize innovation. The author hopes for a dot‑com–style reset: a market correction that toppled some overhyped companies but left durable infrastructure and a culture of hands‑on tinkering. They introduce a playful metric, the C/B ratio (conferences-to-blogging), to signal when a technology has normalized: lots of blogging and small open projects, not perpetual glossy conferences. That normalization matters because it shifts AI from VC‑driven hype and centralized control (OpenAI, Nvidia, Google) toward grassroots engineering, long-form technical writing, and messy but fertile open‑source experimentation. Technically and socially, the article warns of mixed consequences. A crash would expose the fragility of an ecosystem hung on a few anchor firms, reduce available funding, and force people to optimize for cost, deployability and guardrails — e.g., running smaller models locally, shipping robust RL/ops practices, and teaching non‑experts safe LLM usage. It also raises safety and societal risks: mass automated content, climate and cognitive impacts, and geopolitical tensions. Still, the author celebrates the upside: cheaper compute and less money chasing every idea could produce more weird, useful hacks, better documentation, and a healthier, less monopolized AI culture.