Since Hobbes, states and employers have asked people to be reasoners first. Hobbes himself showed that reasoning is a learned job the state needs, and now that machines can do much of the job, being human may turn out to mean opening a path toward transcendence, the way James Turrell’s rooms of light do.
The notion of cognitive offloading, handing mental work to a tool so that attention can go somewhere else, is now central to the debate over AI’s impact on society. Researchers at MIT, Microsoft and Carnegie Mellon are measuring what it costs students and knowledge workers, and every enterprise AI rollout is a bet on what it returns. Followed back far enough, the trail runs centuries past the technology, and one of its clearest early markers is Thomas Hobbes’s Leviathan, published in 1651. Hobbes asked what reason is, where it comes from and what the state needs from it, and AI researchers and executives are asking the same questions again now that machines are in the room.
For most of modern history since Hobbes, people living inside states and companies have understood themselves first as reasoners: people who keep accounts, weigh consequences and explain themselves in terms someone else can check. Hobbes showed that this kind of reason is a skill people learn because the state needs it, which is a long way from proof that reasoning is the whole of a human being. Machines can now carry a large share of that reckoning. With the reckoning handled, people can take seriously the possibility that the human task was always something else, closer to what the artist James Turrell builds his rooms of light to do: opening a path from ordinary perception toward something larger than it.
Hobbes Drew the State From the Inside and the Outside
Hobbes was a chronic ontologist, a man who could not stop asking what a thing is made of, whether the thing was a body, a thought or a government, and he lays out the method he used on the state plainly in the Introduction to Leviathan. Its motto is “Nosce Teipsum, Read Thy Self”: whoever looks into himself and considers what he does when he thinks, reasons, hopes and fears “shall thereby read and know, what are the thoughts, and Passions of all other men.” Reading other people by their actions alone, Hobbes warns, is “to decypher without a key,” and it only ever covers a person’s acquaintances, “which are but few.” A ruler has to read mankind in himself and then check that reading against what people do in the world.
Readers who have never opened Leviathan can follow Hobbes by following that double reading. Looking inward, he found the parts of a person, which come down to passions such as fear and a capacity to reason that has to be trained. Looking outward, he found what those parts produce when millions of people share a territory, and he drew the edges of the state where the two readings met. The opening image of Leviathan compares the heart to “a Spring,” the nerves to strings and the joints to wheels, and then describes the commonwealth as an artificial man with law as its “artificiall Reason and Will.” Hobbes treated the state as a machine built out of the same parts he had found in himself.
Reason Is Reckoning, Learned Like Bookkeeping
In Chapter V Hobbes defines reason as “nothing but Reckoning (that is, Adding and Substracting)” of the consequences of names people have agreed on. People are not born with it the way they are born with sight and memory, he writes, because it is “attayned by Industry,” learned by getting words straight and following a method. Descartes had made thinking the proof of his own existence about a decade earlier. Hobbes put a large part of thinking in the same category as bookkeeping, and on his account the only question worth asking about a machine that does some of the reckoning is whether it got the sums right.
The State Rewards the Reasoner
To be taxed, hired, paid and governed, a person has to keep names straight, balance the books, and give explanations that hold up after he leaves the room. Hobbes never says the state invented reason, and people reasoned long before governments existed. Governments and companies reward the particular kind of reasoning that turns what happened into a line someone else can audit, and marketers, analysts and strategists spend whole careers producing it. That work has value, because a promise nobody can interpret is worthless and a public decision nobody can question is dangerous. After four centuries of being paid and judged for it, though, people have come to treat a requirement of the system as the definition of who they are.
Some Thinking Changes a Person and Some Is Just the Job
The MIT Media Lab had students write essays with ChatGPT over four months and found them falling behind on brain activity, language and behavior, and the researchers called the result cognitive debt. Microsoft Research and Carnegie Mellon surveyed 319 knowledge workers and found that the more they trusted AI, the less critical thinking they reported doing. Both research teams count every step a person stops performing as a loss. That holds for a student learning to write, and it holds much less for an executive on the fourth round of redlines on a vendor contract, who wants the contract signed.
Farmers who gave up the plow kept the land and the craft, and gave up the duty to pull. They wanted the harvest, and the strain on their backs was what it cost. Reading a difficult book or working through a proof changes the person doing it, and that kind of thinking is worth keeping. Reconciling accounts, or holding every loose end of a project in one head because the software can’t, is reckoning the system demands, and by Hobbes’s own definition a machine can do it.
Hand Off the Reckoning and Keep the Responsibility
Hobbes saw the risk in the same chapter. He describes a head of household who adds up all the bills into one total without checking any of them, and says that man might as well have accepted the account on faith, because whoever takes conclusions on trust “does not know any thing; but onely beleeveth.” The argument of The Rear-view Odometer And Its Antidote was that once execution is cheap, people get paid for judgment and for owning the result, and Hobbes’s careless householder is that argument made in 1651. A person can offload safely only when he can see where an answer came from, correct it or refuse it, and find out what happened after it was used. Fei-Fei Li sets the same standard in her interview with Issues in Science and Technology: “AI should improve the human condition.”
Turrell Opens a Path Past Reckoning
James Turrell’s Life Work, Rodin Crater. Source: Nevada Museum of Art
James Turrell builds rooms for looking at light, and he traces the work to his Quaker grandmother, who told him to “go inside and greet the light,” according to Smithsonian. In 2000 he built a version of that instruction for the Live Oak Friends Meeting House in Houston, a Quaker meeting house with a roof that opens so that, in his words, “the sky is really brought down to you.” Quakers worship in silence, waiting for what George Fox called the inner light, and Turrell gave that waiting a room. “I want you to sense yourself sensing. To see yourself seeing,” he told Smithsonian, and he describes his medium as perception itself.
Turrell and his engineers do an enormous amount of calculation to build each of those spaces, and a visitor sitting inside sees none of it. “It’s one thing to know the math,” Turrell said of Roden Crater, his observatory in the Arizona desert, “but I want you to feel the shape change as a real, physical experience.” In Turrell’s rooms Hobbes’s reckoning goes into the construction, and the person in the room is left free for an experience no account of the construction could replace. Machines can now take over the reckoning in ordinary life the way Turrell’s engineers take it over in his rooms.
Hobbes built his artificial man to protect the natural one, “for whose protection and defence it was intended.” The companies building artificial intelligence should be asked whether their machines can do something Hobbes never asked of the state, which is to give people back the room the reckoning took. The hours a model frees can just as easily go to more reports and a faster version of the same routine, and an employer measuring output will claim them first. Spent the other way, those hours belong to a task far older than the state, the one Turrell’s grandmother named: going inside to greet the light. Farmers kept the harvest when they gave up the plow, and for people released from four centuries of reckoning, the harvest may be transcendence itself.
Alan Eyzaguirre writes about AI, capital markets, and corporate strategy as a Silicon Valley entrepreneur and corporate strategist.