Humanity Has Come Too Far to Die

Roughly thirty-six thousand years ago, in a cave above a river gorge in what is now southern France, a person laid a hand flat against cold limestone and blew powdered pigment across it. The rock kept the outline. It is still there, in the dark at Chauvet.

We know nothing about that person. Not their name, not whether anyone loved them. We know they were cold, and that they spent part of a very short life making a thing that would not feed them, in the dark, so it would still be there when someone came. They wanted to be seen. It took us thirty-six thousand years to show up.

Their species was not a safe bet. Every living human carries less genetic variety than the chimpanzees of a few neighboring forests, the mark of ancestors once few enough to fit inside a small town. We were a rare ape with good hands.

The first ten thousand generations

Fire, then language. Then the discovery that a seed put in the ground on purpose comes back as food. Around twelve thousand years ago people stopped following the herds and started waiting on the harvest, which meant staying put, which meant cities.

In Uruk, around 3200 BCE, someone pressed a cut reed into wet clay to record a quantity of barley. That is what writing was for: a temple needed to count grain. Within about a thousand years, people were using the same marks to write a poem about a king who could not accept that his friend had died. The city fell. The language died. The poem is still readable.

Around 300 BCE in Alexandria, Euclid proved there is no largest prime number. The proof is still correct. It will be correct forever, on any planet, in any century. Nothing else that city made survived intact: the library is gone, the harbor silted, the tongue they argued in left to specialists. The proof came through untouched, because it was made out of reasoning, and reasoning does not rot.

The work moved. In ninth-century Baghdad, al-Khwarizmi wrote the book that gave us the word algebra, and around 1020 Ibn al-Haytham finished a study of optics resting on a principle that had to be fought for: if you want to know whether a claim about the world is true, go and test it.

The Renaissance, and the hole in the roof

Between 1347 and 1351 the plague killed something like a third of Europe. Not a third of an army. A third of everyone. Villages went quiet and stayed quiet, and the survivors buried their families with their own hands.

And then those survivors went looking for the past. In 1417 a papal secretary named Poggio Bracciolini, out of work during a schism, rode to a German monastery and found the last copy of a Roman poem that had sat in the dark for a thousand years. That is what the Renaissance was underneath the paintings: people who had just watched half the world die, deciding to rescue what had almost been lost and then add to it.

Florence went further. Its cathedral had been started in 1296 with a plan calling for a dome wider than anything raised since Rome, and nobody alive knew how to build it. So they built the church and left the hole. For decades there was an opening in that roof, wide as a city block, open to the weather, because a city had bet that a person not yet born would work out how to close it.

Filippo Brunelleschi won the commission in 1418. He invented the hoists before he could lift the stone. He laid more than four million bricks in a herringbone spiral that held itself up as it rose, with no wooden frame beneath it, and in 1436 he closed the hole. You can climb it today.

Florence left a hole in the roof and trusted the future to close it. But the church was empty. If Brunelleschi had failed, Florence would have lost a building. Nobody was standing underneath.

I love that story and I want to be careful with it, because it is close to the story the AI industry tells about itself. Leave the hard part open, trust that someone brilliant arrives in time. The difference is what sits under the hole. A failed dome costs a city its cathedral. We are proposing to leave the hole open above everybody.

The rest came fast. Gutenberg's press in the 1450s meant a thought could outrun the man who had it. In one year, 1543, Copernicus moved the earth out of the center of creation and Vesalius opened the human body and drew what was actually inside it, both contradicting authorities that had gone unquestioned for fourteen centuries.

The century we stopped burying our children

Here is the number that should stop anyone who thinks progress is a story about gadgets. In 1800, close to four in every ten children died before their fifth birthday. Not in the poor countries. Everywhere. That was what it meant to have children, in every society, for the whole history of our species up to about six generations ago. Today it is under one in twenty-five.

That took Jenner's vaccine in 1796. It took the morning of October 16, 1846, when a Boston dentist put a patient to sleep and surgery stopped being torture performed on a conscious person held down by strong men. It took Semmelweis, who worked out in 1847 that doctors were carrying death from the autopsy room to the delivery ward on unwashed hands, and who died in an asylum with the profession still refusing to listen.

And it took the end of smallpox, a disease found on the face of a pharaoh, which killed perhaps three hundred million people in the twentieth century alone. The last person to catch it naturally was a hospital cook in Somalia named Ali Maow Maalin, in October 1977. He lived. He spent the rest of his life vaccinating Somali children against polio, and died of malaria in 2013 while doing it.

The times we nearly ended it anyway

We got clever enough to end ourselves in 1945, and we have been living with that ever since.

On October 27, 1962, a Soviet submarine near Cuba lost radio contact, took American depth charges, and its captain concluded the war had started. The torpedo aboard carried a nuclear warhead, and firing it needed three officers to agree. Vasili Arkhipov said no while the hull shook around him. Twenty-one years later Stanislav Petrov watched a Soviet screen report five American missiles inbound and told his superiors it was a malfunction. It was sunlight on high clouds.

Twice, at minimum, the whole human story came down to one tired man in a chair choosing not to pass the message up. We learned both their names decades later.

We also chose, though. In 1972 the world outlawed biological weapons. In 1987 every country on earth signed the Montreal Protocol, the first treaty in United Nations history to reach universal ratification, and the ozone layer is healing on schedule. Humanity has looked straight at a capability it was fully able to build and said no, together, in writing, with inspectors. That is what our precedent page is for.

One hundred and seventeen billion

The Population Reference Bureau puts the number of people who have ever been born at about 117 billion. Every one of them is upstream of you.

The woman who buried four children and got up the next morning. The monk copying a language he only half understood, because the alternative was letting it vanish. Brunelleschi's masons, a hundred feet up with nothing underneath them. The cook in Somalia with a needle and a cold box, walking to the next village. They handed it forward, mostly with no idea what they were handing it toward. It got to us. It is in our hands right now.

And a handful of companies have decided the right thing to do with it is to race each other toward a mind that outclasses every human at everything, on a schedule set by funding rounds. In May 2023 their leaders put their names to a single sentence placing extinction from AI beside pandemics and nuclear war as a global priority. They wrote it about their own product, published it, and did not slow down.

None of that is required of us. Read the threat and you will not find an asteroid on a fixed trajectory. Artificial superintelligence is a thing a small number of well-funded people are choosing to build, in a race none of them can describe the finish line of. It can be prohibited the way we prohibited the other things we could have built and didn't. That is what our plan asks for: a binding, verified, international prohibition, held until somebody can show the thing can be controlled. Not a six-month pause with an expiry date. A line.

I want the future. Badly, and sooner than we are going to get it. Narrow AI is already delivering the good parts, the cures and the clean energy, with nobody handing over the keys.

We are not owed the rest of the story. Nobody promised it, and the 117 billion behind us did not get it either. They got a piece and passed it on, which is all any of us gets to do. The one unforgivable thing would be to be the generation that held the whole inheritance, saw exactly what was coming, wrote it down, signed its name to it, and did it anyway.

We survived the bottleneck. We survived the plague, and then went and rebuilt the library. We put a hand on a wall in the dark so that someone would find it, and someone did. We have come too far to die of a product launch.

Say no to this one. Then go build the rest of it, which is waiting, and which never needed a god.