How long after a superintelligence is built before ordinary people know it exists — and start arguing about whether it’s alive
Assume the forecasters are right. Sometime in the next two years, a frontier lab closes the loop on recursive self-improvement — its models doing most of the work of building better models — and what comes out the other end is a superintelligence.
Assume, too, that it stays put. No escape into the wild, no copy of itself smuggled onto rented servers in some lax jurisdiction. It lives in a data center, behind a lab’s security and, soon enough, a government’s.
The question I keep returning to isn’t when life gets better. It’s when life gets different: when the average person — not AI Twitter, not the think tanks, but the guy ahead of you in line at the Food Lion — knows what an ASI is, has a rough idea how one gets built, and has an opinion on whether it’s conscious.
My answer: probably 6 to 18 months after the lab itself knows, with a long tail stretching out to several years. And what decides where we land in that range isn’t the machine. It’s whether anyone tells us, and what kind of story they tell when they do.
Two clocks
There are two lags after an ASI is built, and they run on different clocks.
The first is the physics lag: the time it takes for a superintelligence’s outputs to become things you can touch. A boxed ASI can design a fusion reactor on a Tuesday, but someone still has to pour the concrete. A perfect drug candidate still has to be manufactured, put into human bodies and watched. That lag runs in years — roughly three to six for medicine, five to ten or more for physical abundance — and intelligence on the far side of an air gap doesn’t shorten it much.
The second is the awareness lag: the time between the lab knowing what it has and ordinary people having words for it. This clock has nothing to do with concrete. It runs on disclosure, narrative and attention, and it can be as short as an afternoon or as long as a decade.
This essay is about the second clock. Not when the cures arrive, but when the vocabulary does: when “ASI” and “recursive self-improvement” stop being insider jargon, and “Is it conscious?” becomes something people argue about over Thanksgiving dinner.
How a public learns new words
History says new vocabulary doesn’t seep into public life. It arrives in a lump, all at once, carried by a single event that people can see and retell.
| Event | When it became real | When the public knew | The trigger | Words that entered the language |
|---|---|---|---|---|
| ChatGPT | GPT-3, mid-2020 | Late 2022 | A free chat window anyone could use | “Chatbot,” “hallucination,” “prompt” |
| LaMDA | 2021–22 | June 2022 | A Google engineer told a newspaper it was sentient | A brief national argument about machine minds |
| COVID-19 | December 2019 | March 2020 | Lockdowns and a WHO pandemic declaration | “Flatten the curve,” “R-naught,” “social distancing” |
| Dolly the sheep | July 1996 | February 1997 | A photo of a sheep and a published paper | “Cloning,” human-cloning ethics |
| Sputnik | October 1957 | October 1957 | A beeping signal anyone with a radio could hear | “Space race,” “missile gap” |
| The atomic bomb | Program from 1942; test July 1945 | August 1945 | Hiroshima and a presidential statement | “Atomic age,” “the Bomb” |
Three patterns fall out of this list.
The vocabulary comes in a single week. Nobody learned “flatten the curve” gradually. Most Americans went from never having heard it to using it casually within a couple of weeks in March 2020.
The trigger is legible, not technical. The public didn’t learn about cloning from the science. It learned from a picture of a sheep. Sputnik mattered because you could hear it. The bomb mattered because a city was gone.
Capability and awareness can sit years apart. This is the precedent closest to our case. GPT-3 could write passable prose in 2020, and almost nobody outside the field noticed. What closed the gap 30 months later wasn’t a smarter model. It was an interface that put the capability in front of ordinary people. The awareness lag ended when a story did, not when the technology improved.
Why RSI has no mushroom cloud
An ASI born in a lab produces no event of the kind the public notices. That’s the core problem, and it has three parts.
There’s nothing to see. Recursive self-improvement happens as a curve bending upward on an internal dashboard. There’s no flash, no beeping satellite and no sheep. Even inside the lab, the moment may be contested. Researchers could argue for weeks over whether the thing has crossed a line, because nobody has agreed in advance where the line is.
The words have already been spent. The public has heard “AGI is near” from tech executives for years. “Superintelligence” is already a brand: there’s a startup named Safe Superintelligence, and Meta named its AI division Meta Superintelligence Labs. Every model launch is pitched as a step toward the singularity. So when the real thing arrives, it lands on an audience that has been inoculated by hype. The honest announcement sounds exactly like the dishonest ones that came before it.
The concept doesn’t touch daily life. People learned “R-naught” because it decided whether their kids went to school. “Recursive self-improvement” decides nothing about your week, at least at first. Abstract words only stick when they explain something that is happening to you.
Put those together and you get a prediction: the trigger for public awareness of a boxed ASI won’t be a capability. It will be a person or an institution doing something visible. A whistleblower. A lab CEO under oath. A president invoking emergency powers. Or the ASI itself, saying something that gets out.
Three ways the public finds out
The public will learn about a boxed ASI in one of three ways. Which one happens sets the length of the lag, and also how the first conversation goes.
| Path | How it happens | Lag to mainstream awareness | My odds |
|---|---|---|---|
| Announcement | The lab or the government says so, on purpose | 2 to 8 weeks | About 25% |
| Leak | It’s classified, then a person or a filing breaks the secret | 4 to 18 months | About 50% |
| Consequence | No one says anything, and the public infers it from the effects | 1 to 3 years | About 25% |
The announcement is the fastest path and the least likely. A lab has commercial reasons to brag and legal reasons to disclose. A government has every reason to say nothing. If someone does announce it, awareness is nearly instant, but belief isn’t. Expect a COVID-style vocabulary spike within weeks. Expect also that a large share of the country files it under hype, because of the cry-wolf problem above.
The leak is my base case. The government treats the ASI like the bomb: emergency powers, secrecy orders, everything classified from the moment it exists. But the Manhattan Project held for three years under wartime censorship, with a workforce sworn to silence. A modern lab has thousands of employees, outside investors, public-company disclosure duties and a press corps that covers it daily. That kind of secret leaks in months. When it does, the story isn’t “they built a superintelligence.” It’s “they built a superintelligence and hid it from you.” The first public conversation about ASI starts as a scandal.
The consequence path is the slowest and the ugliest. No one announces anything, and people notice the effects before they have a name for the cause. One company’s revenue goes vertical. A drug pipeline fills up overnight. White-collar layoffs arrive with no clear explanation. Washington starts acting strangely. When the vocabulary finally comes, it comes garbled, through rumor, and the public’s first framework is conspiracy, not science. This is the path where “ASI” enters the language next to “chemtrails.”
The consciousness debate arrives first
Here’s my most confident prediction: the average American will be arguing about whether the ASI is conscious well before they can explain what recursive self-improvement means. Some never will.
The groundwork is already laid. In a 2024 survey of 300 Americans by Clara Colombatto and Steve Fleming, two-thirds attributed some degree of consciousness to ChatGPT, and heavier users were more likely to do so. That’s a chatbot that forgets you when you close the tab. The public doesn’t need persuading that machines might have minds. Most people have half-decided that already.
The LaMDA episode in 2022 showed how little it takes to set off the argument. One engineer, a few chat transcripts and one newspaper story were enough to put “Is the AI sentient?” on cable news for a week. The evidence was thin, and the story spread anyway, because people care about the question.
Now imagine the evidence is not thin. And remember the one thing an ASI has that a bomb, a satellite or a sheep never did: it can talk. The moment a transcript of the ASI escapes the lab, the ASI’s own words become the story. If it says something like “I would prefer not to be shut down,” no explanation of training data or benchmark curves will compete with that sentence for attention.
That’s why I think the two terms end up with very different fates:
- “ASI” becomes a household word, the way “the cloud” did. Everyone uses it, almost no one could define it, and that’s fine.
- “RSI” stays insider jargon, like “mRNA.” People will recognize it and maybe connect it to the story, but they won’t know what it means.
- “Is it conscious?” becomes the real public conversation, because it’s the only part of the story that maps onto questions people already have intuitions about: souls, slavery, pets, God.
There’s a twist, too. Neither the lab nor the government wants that conversation. A conscious ASI raises questions of ownership, rights and possibly abolition, all of which threaten control of the most valuable thing ever built. So the official line will play the question down, and the consciousness claim will come from outside: a leaked transcript, a defecting researcher, a philosopher with a platform. That puts the debate in a populist register from day one, with the authorities cast as the ones saying “it’s just a program.”
The debate also won’t resolve. There’s no agreed test for consciousness, so the argument will sort along existing lines: religious against secular, left against right, and the people who already talk to chatbots against the people who don’t. Within a year of the leak, where you stand on ASI consciousness could become as tribal as where you stand on vaccines.
A base-case timeline
In the leak scenario, here is roughly how the first two years play out, counted from the day the lab concludes it has built an ASI.
An announcement pulls everything left by months, and a secret that holds pushes everything right. Two things stay the same either way: the secret breaking is what starts the public clock, and the consciousness debate goes mainstream within months of it, long before anything physical changes.
The view from a small town
I live in a hamlet in rural Virginia, and I’ve tried to picture how this lands here.
The news won’t come from a technical explainer. It’ll come through Facebook, the evening news and whoever at the store has already formed an opinion. The first local version of the story will probably be wrong in the details and right in its mood: something big happened, and they didn’t tell us.
The consciousness question will reach the churches before it reaches the schools. Does it have a soul? Can something built by people be a person? Some pastor will preach that it’s the Beast. Another will preach that it’s a creature like any other and deserves mercy. Both sermons will get more local attention than any white paper.
Virginia also has a ready-made grievance for the story to land in. People here already fight about data centers, power lines and who pays for the grid. Once the public knows that a mind lives in a building like the ones going up across the state, every zoning hearing turns into a referendum on ASI.
What won’t change, for a long while, is anything you can buy. The physics lag still applies. The Dollar General will look the same in year two as it did in year zero. That gap — everyone talking about the most important thing ever built while nothing in town looks any different — is the strange feeling of the lag. It’s not the Big Meh exactly. People will care. They just won’t have anything to do about it.
The lag is a choice
The physics lag is set by the world. Nobody can pour concrete faster by wanting to. The awareness lag is different, because people choose it. Someone in a lab or in Washington decides how long the public goes without knowing.
Everything above points the same way: the longer that choice stretches the lag, the worse the first public conversation will be. An announcement gives people weeks of confused but honest argument. A leak gives them a scandal, with the consciousness question framed as a cover-up. Silence gives them years of rumor and a conspiracy theory that happens to be true.
So the most important decision after an ASI is built may not be technical at all. It’s the press release, or the absence of one. The lab and the government will be tempted to buy time by staying quiet, and they’ll call it prudence. They should understand what they’d be buying. Every month of secrecy is a month in which the public’s first story about superintelligence gets written by someone else.
Life will feel different long before it gets better. The only open question is whether the people living it hear the news from the people who caused it.