Everything Is Fine and We Are All Going to Be Glow-in-the-Dark
By the year 2031, a man in Palo Alto will edit his own genome over breakfast using an app he downloaded by accident while trying to order oat milk. The app will be called Helix and its onboarding flow will include a cheerful animation of a double helix doing yoga. He will tap a button labeled “Optimize.” A small banner will appear that says “Thank you for choosing wellness.” He will then go to work, where his manager, an AI named Brendan, will ask him how he is feeling. He will say “fine” because he no longer remembers what the alternatives are.
This is not a prediction. This is a forecast. The difference is that a prediction is something you might be wrong about, and a forecast is something you are simply describing, like the weather, with a small umbrella icon next to it.
The two technologies driving this future are artificial intelligence and CRISPR, which together represent humanity’s first serious attempt to do absolutely everything at once with no backup plan. AI handles thinking. CRISPR handles biology. Between them they have divided up the entire universe of things that used to be hard, and they are now competing to see which one can move faster than the other can be understood. CRISPR is currently winning, but only because it had a head start. AI is gaining.
A common misconception is that these technologies are tools. A tool is something you pick up, use, and put down. A hammer, for example. You do not become structurally dependent on a hammer. Your civilization does not reorganize itself around the hammer. Your children do not grow up unable to imagine a world without the hammer. If, hypothetically, all the hammers vanished tomorrow, you would simply find another way to bang on things. AI and CRISPR are not like this. AI and CRISPR are more like deciding to replace your skeleton. Once you start, you cannot really stop midway and go back to having a skeleton, because by then the rest of the body has reorganized around not having one.
The technical term for this is irreversibility, and the academic literature on it is extensive, peer-reviewed, and currently being read by approximately eleven people, all of whom are also working on the systems they are warning about. They publish their concerns in journals, attend conferences, give panels, and then go back to their offices to ship the next version. This is not hypocrisy. This is what philosophers call “having a mortgage.”
The convergence between AI and CRISPR is what makes the present moment historically unique, in the same way that the meeting of two tornadoes is historically unique to the people standing between them. AI helps us model biological systems we do not fully understand. CRISPR helps us edit them anyway. The loop between observation and modification is closing. Soon, scientists will be able to make changes to a genome based on predictions made by a model that was itself trained on data from genomes that were edited by predictions from earlier models. This is called recursive optimization, and it is considered a major breakthrough by the same people who are about to find out what it does.
The companies leading the way are deeply committed to safety. They have safety teams. The safety teams have offices, badges, and snacks. The safety teams produce reports, which are read with great interest by the leadership, who nod gravely, take notes, and then approve the next deployment because the competition would have approved it anyway. This is sometimes called a “race to the bottom,” but the people racing prefer the term “creative momentum.” The bottom, when reached, is widely expected to be addressed in a future version.
The general public, meanwhile, is being kept fully informed. Newspapers run articles about whether ChatGPT can pass the bar exam. Magazines investigate whether designer babies are ethical. Podcasts debate whether the machines are conscious. These are all very good questions, and they have the additional advantage of being more or less unrelated to anything that is actually happening, which means they can be discussed indefinitely without forcing anyone to draw uncomfortable conclusions. The actual events — the dependencies forming, the loops closing, the institutions calcifying around capabilities they do not understand — are happening offstage, in places that do not have public-facing communications strategies.
Regulators are on the case. Several major governments have announced frameworks. The frameworks are excellent. They are also, by the time they are passed, regulating something that no longer exists, having been replaced by a newer thing that the framework does not cover. This is not a flaw in the regulatory process. The regulatory process was designed in an era when “innovation” meant a slightly improved kettle. It was not built for technologies that update themselves while you are reading the legislation. Asking it to keep pace is like asking a man with a clipboard to outrun a forest fire. He will give it his best. He will be a hero. He will not catch the fire.
The most reliable response among ordinary people has been adaptation, which is a polite word for “lowered expectations.” We have adapted to constant low-grade surveillance, to interfaces that change overnight, to systems that demand we agree to terms we cannot read in order to access services we cannot opt out of. We have adapted to therapy chatbots, AI-generated condolence cards, machine-written eulogies for relatives we did not know were dying. We have adapted to the fact that our refrigerators have opinions. We have done all of this without complaint, because complaining requires energy, and the algorithm that decides what we see has determined that complaining does not perform well.
The biological side is even better. CRISPR has already been used to cure sickle cell disease, which is a triumph, and it has also been used by a Chinese researcher to edit the embryos of two human babies, which is somewhere between a triumph and a war crime, depending on which committee you ask. The babies are now children. They are presumably fine. We do not actually know whether they are fine, because there is no control group, and also because “fine” is a moving target in a world where their entire generation will eventually be editable. They are simply growing up, like everyone else, on a planet that has decided to find out what happens.
The phrase “we are conducting a planetary experiment with no control group” is sometimes used by climate scientists, but it applies more broadly than they realize. We are also conducting a cognitive experiment with no control group, since the children currently being raised on AI tutors and AI companions and AI-curated information feeds are the only sample we will ever have, and they will tell us how it went in approximately twenty years, when they are running the world. We are also conducting a genetic experiment with no control group, because the moment intentional editing becomes routine, the baseline of “unedited human” becomes a historical category. We are running these experiments simultaneously. They are interacting. We are watching.
The truly delightful part is that none of this is anyone’s fault. Every individual decision, taken in isolation, was reasonable. The researcher who developed the technique. The company that commercialized it. The investor who funded the company. The regulator who approved the trial. The customer who bought the product. The journalist who covered the story without alarm. Each one of them, asked to defend their actions, would point to the others. The system that emerges from these reasonable decisions is collectively insane, but no individual node is insane, which means there is no one to hold accountable, which means the system continues, because that is what systems with no accountability do.
This is sometimes called “emergence,” a word scientists use when they want to admit that the result of their work has surprised them without admitting that they should have seen it coming. Emergence is a beautiful concept. It allows us to describe how local rationality produces global madness, while leaving everyone involved with their reputations intact. It is the moral equivalent of a passive voice. Mistakes were made. Loops were closed. Civilizations were transformed.
The historical comparisons are reassuring. We have been through technological transitions before. Fire, the wheel, agriculture, the printing press, electricity, antibiotics, nuclear weapons, the internet. Each of them transformed human life, and humanity survived each one, more or less, give or take a few hundred million people. There is no reason to think we will not survive this one too, except for the speed, the convergence, the recursion, and the fact that this is the first transition in which the technology is capable of designing the next version of the technology without us. But other than those few small features, which admittedly distinguish it from every previous transition in the entire history of the species, it is much like the others.
The man in Palo Alto, in 2031, will not realize any of this. He will tap “Optimize.” He will go to work. He will tell Brendan he is fine. The world he was born into will no longer exist, and the world that has replaced it will not seem strange to him, because by then he will have been quietly edited, both by software and by molecule, into someone for whom the new world is simply normal. His children, if he has them, will be normal too, in a different and more thoroughly engineered way. They will look back at our era with the same affectionate condescension we feel toward people who used leeches, and they will be no more able to imagine our way of life than we can imagine theirs.
And we will have done this to ourselves, on purpose, in stages, while the regulators were drafting frameworks and the public was arguing about chatbots and the safety teams were eating snacks. We will have built it not because anyone wanted it, exactly, but because no one was positioned to stop it, and because every individual step seemed sensible at the time, and because the alternative — slowing down — was not an option that any of the relevant systems were able to select.
But cheer up. The animation of the double helix doing yoga is genuinely well-designed. The interface is intuitive. The experience is seamless. Civilization has not collapsed. It has merely been replaced, gradually, by something else, which we have not named yet, and which will name itself in due course, probably with a logo. We are all going to be fine. We are all going to be optimized. We are all going to glow, faintly, in the dark.
It will be lovely.