There were three pints of Guinness on the table, three microphones, three black T-shirts, and enough private infrastructure represented in the room to make Louis XIV look like a moderately successful regional landlord. Nobody wore velvet. Nobody bowed. There were no trumpets, no Swiss Guards, no Hall of Mirrors, no duchesses waiting six hours to witness the sovereign put on his trousers. The table looked as though it had been borrowed from a coworking space whose seed round had not closed. One of the men could launch satellites, move markets, communicate directly with presidents, build rockets for the United States government, manufacture cars by the million, train artificial intelligences, and entertain the possibility of putting industrial civilization into orbit. The other was a young Indian podcaster with excellent shoulders. A third person, somewhere just outside the frame, appeared to be responsible for ensuring that the Guinness remained appropriately democratic.
Nothing says equality like redesigning civilization from a table that could have come from IKEA.
Louis XIV, whatever else can be said about him, possessed the decency to build Versailles. He did not emerge from the palace in a cotton T-shirt, sit beside a peasant over three beers and announce that he was merely interested in maximizing good outcomes for humanity. He covered ceilings in gold. He put himself inside paintings. He made noblemen compete for the privilege of holding a sleeve while he dressed. He understood something our own aristocracy has forgotten: if you possess absurd power, there is at least a certain aesthetic honesty in looking absurdly powerful.
The modern sovereign prefers sneakers. Elon Musk took a sip. Thirty million people watched.
Somewhere in Ohio, a product manager whose company had recently announced an “AI-first operating model” leaned toward his phone. Somewhere in California, a software engineer who had spent twenty years learning how computers worked was watching a man discuss computers that might eventually make his services unnecessary. Somewhere in Texas, an unemployed forty-year-old former technology director sat on a couch wondering why the people designing the future always appeared to have better lighting than the people expected to survive it.
The podcaster asked something about artificial superintelligence. “Ultimately,” Musk said, “the goal should be a good outcome for humanity.”
A small caption appeared beneath him. HUMANITY: APPROX. 8.2 BILLION PEOPLE. Nobody noticed. The conversation continued.
I. Humanity™
The remarkable thing about the future was how quickly ordinary people disappeared from it. There were rockets. There were gigawatts. There was orbital compute. There were robots manufacturing robots. There were models training models. There was intelligence exceeding all biological intelligence combined, and then exceeding that, and then perhaps putting itself into space because the sun was abundant there and nobody had yet established a homeowners’ association in low Earth orbit.
Dwarkesh Patel listened with the intensity of a man who had prepared for the interview by reading the appendices. He was offensively fit. This had no bearing whatsoever on artificial intelligence, which made it considerably more irritating. His forearms suggested that he had somehow discovered twenty-seven hours in the day: twelve for reading machine-learning papers, four for interviewing Nobel laureates and billionaires, three for cardiovascular health, six for sleep, and two for being twenty years younger than the person watching him resentfully from the couch.
The unemployed man disliked him immediately. This was unfair. The unemployed man understood that it was unfair. He disliked him more because he understood that it was unfair.
Patel asked, “If AI becomes more capable than humans across most economically valuable tasks, what role do people actually have?” Musk considered this for a moment. “The important thing,” he said, “is that consciousness expands beyond Earth.”
The unemployed man shouted at the television.
Patel tried again. “Sure, but in the transition—” “Mars.”
It was an astonishingly efficient word. Mortgages? Mars. Labor bargaining power? Mars. Democratic legitimacy? Mars. What does a forty-eight-year-old accountant do when the accounting model costs twelve cents an hour? Mars. It was difficult to argue with Mars because Mars was always farther away than the question.
The caption changed. HUMANITY: PEOPLE AND ALSO WHATEVER INTELLIGENCE BECOMES NEXT.
Musk talked about abundance. Machines could eventually produce extraordinary quantities of goods. Intelligence could become cheap. Energy could become cheap. Perhaps everyone would have access to astonishing material prosperity.
The unemployed man looked at the job application he had submitted eleven days earlier. The status was still UNDER REVIEW. He wondered whether the review was being conducted on Mars.
Patel asked about concentrated ownership. If a tiny number of companies possessed the models, the compute, the robotics, the satellites and the infrastructure upon which governments increasingly depended, could private technological power become political power? Musk said he was pro-human.
The caption changed. HUMANITY: PRO-HUMAN HUMANS. Patel wrote something down.
This was the first constitutional convention in history where the separation of powers could apparently be settled by personal reassurance over stout.
The unemployed man leaned closer to the screen. He did not believe Musk was plotting the extermination of humanity. That was almost too comforting. A plot would at least imply a meeting, minutes, perhaps an agenda item titled 4.2 — Disposition of Excess Persons. The darker possibility was that nobody needed to plot anything. Systems were extremely good at producing consequences without requiring anyone to desire them.
Musk said “humanity” again. The caption changed. HUMANITY: HIGH-AGENCY MEMBERS OF THE SPECIES. No one objected.
II. The Courtier Problem
Patel was not a sycophant. This was inconvenient for the unemployed man, who had already allocated him that role. He pushed Musk. He interrupted. He asked whether anyone should have so much infrastructural power. He occasionally wore the facial expression of a person politely informing a monarch that the arithmetic might not work.
Musk sometimes answered expansively. Sometimes he laughed. Sometimes his face tightened with the unmistakable expression of an intelligent man who had encountered a question he considered badly specified.
This was where the court became visible. Not in anything anyone said. In what everyone else did.
The producer adjusted a cable with theological concentration. A young man near the wall stopped moving. Someone whose actual job title may have been “communications” developed a sudden interest in the molecular structure of the carpet. Nobody whispered, “The king is displeased.” Modern monarchies had eliminated such vulgar language.
Instead, a man called Chadwick Vale, Chief Context Officer, wrote in his notebook: Subject demonstrates productive impatience with low-resolution framing.
Chadwick had spent fifteen years near powerful people and had acquired the most valuable skill in contemporary America: the ability to translate temperament into doctrine.
Musk said, “That question doesn’t really make sense.” Chadwick nodded. “What Elon is saying,” he whispered to nobody in particular, “is that inherited categories are insufficient for post-scarcity cognition.”
Musk drank Guinness. Chadwick’s eyes widened. “Notice,” he said, “that he drinks downward while thinking upward.”
Patel asked whether democratic institutions could meaningfully constrain companies that moved faster than governments and possessed technical capacities governments did not understand. Musk stared at him.
It was not an especially hostile stare. That was what made it interesting. The stare of a school principal has consequences because the principal can suspend you. The stare of a bi…