On: Bill Gates says we’ve passed AI’s danger thresholds. Now what?

A wealthy man gazes at blue water and announces that the danger thresholds of artificial minds lie behind us. Behind us! This ordering offends every instinct I possess. A threshold is a specification, written down before the apparatus is energized. To cross one unawares and then inquire “what now” is to confess the laboratory kept no instruments on the panel.

Begin at the ceiling, as I always do: a machine intelligence at its theoretical limit could balance every generator on a continent against every load, fold proteins by the thousand, waste not one joule between plant and lamp. That future sits perhaps two decades out. Between here and there lies precisely nothing but the question this man forgot to ask before the crossing: who holds the governor?

Everyone debates how clever the machine will become. Wrong variable. Cleverness is potential energy - inert until coupled. The overlooked mechanism is the coupling itself: how directly the thinking attaches to the switching. A vast intellect sealed in glass can only contemplate. A modest intellect wired to the breakers can darken a coast. Danger resides not in cognition but in actuation, and actuation is architecture, entirely within our hands.

The tractable path is dull, which explains why it waits: mechanical interlocks on everything physical, records kept at the speed of the relay, responsibility pinned to whoever closes each circuit. No law of nature forbids a single item. The absence of these items tells you the failure was never technical.

So when he asks what now - install the gauges retroactively, and accept the shame owed to any builder who constructs the engine first and invents the governor afterward. Resonance rewards preparation. It punishes celebration.