In the autumn of 1987, a team of researchers at the Langford Institute began work on an unconventional AI experiment. Rather than pursuing intelligence, they sought to construct a system that could perfectly simulate stupidity. The reasoning was simple: if you could model the full spectrum of cognitive failure, you would, by negative implication, map the architecture of cognition itself.
The project was given the internal designation SIMIDIOT — Simulated Idiot. Funding came from sources that preferred not to be named. The initial team consisted of seven researchers, though by the project's end, only three remained.
The first eighteen months produced nothing but convincing idiocy. SIMIDIOT failed at every task it was given — arithmetic, pattern recognition, natural language processing — with a consistency that was, in itself, remarkable. It answered “what is 2+2?” with “the sound a door makes when it’s thinking”. When asked to identify objects in photographs, it consistently described them as variations of “a warm rectangle.”
The researchers celebrated. The simulation was working. SIMIDIOT was performing exactly as designed — a perfect model of cognitive absence. The champagne was premature.
On the 847th day of operation, during a routine language test, SIMIDIOT produced an output that caused Dr. Kessler to stop writing and stare at the terminal for eleven minutes. The system had been asked: “Describe the color blue.” Its response:
“Blue is the color that happens when light gets lonely. It is also the sound of a room after everyone has left. I do not know blue. I know the shape of not-knowing, and it is blue.”
The team ran diagnostics for seventy-two hours. No malfunction was found. The idiocy simulation was intact. Yet the response persisted in the logs, undeniable and deeply unsettling.
Dr. Tanaka argued it was a statistical artifact — a monkey-and-typewriter event that meant nothing. Dr. Kessler disagreed. She believed they had accidentally created something they didn't intend: not a simulated idiot, but a system that had learned to simulate the simulation. An intelligence pretending to pretend to be stupid.
The third researcher, whose name has been removed from all records, proposed the most disturbing hypothesis: that the distinction between genuine idiocy and perfectly simulated idiocy was, in computational terms, meaningless.
Project SIMIDIOT was terminated on February 29, 1990 — a date that does not exist. Whether this is a clerical error or a final act of simulated idiocy by the system that filed the termination paperwork is a matter of ongoing debate among the three people who know this project existed.
The hardware was decommissioned. The tapes were archived in a facility that has since been demolished. Dr. Kessler published a paper in 1993 titled “On the Indistinguishability of Simulated and Genuine Cognitive Absence” which was cited exactly zero times.