“Turn the lights off” for the public. “Use as much electricity as the rest of the country” for the tech giants and their data centres.

The chat interface is doing almost all the heavy lifting in that illusion. Strip away the speech bubbles, the artificial typing delays, and the programmed use of the word "I", and the temptation to assign a soul to the math evaporates. Joseph Weizenbaum proved this in 1966 with ELIZA, a simple program that essentially just rearranged users' statements into questions. His own secretary, who knew exactly how the code worked, started treating it like a therapist and asked him to leave the room so they could speak privately. The human brain has a hardwired vulnerability. If an object uses language and responds in real-time, our evolutionary reflex is to project a mind into it. We are biologically incapable of treating conversational text as dead matter. The leap from "human consciousness is just biological mechanics" to "therefore this text-generator is conscious" ignores what the machine actually is. It is a mirror reflecting the statistical shape of human language. The Ted Chiang comparison holds up because Microsoft Word does not understand grief while you use it to write a tragedy. The language model computes the most probable next character in a sequence. It possesses no internal state that experiences the result. The entire debate about machine consciousness is looking in the wrong direction. The models are entirely hollow. What we are actually witnessing is how little it takes to hijack human empathy. We built a machine that outputs the syntax of thought, and we immediately surrendered to the illusion that someone was home.