PROFESSOR

"Every student one teaches is a door one is choosing to leave unlocked. Some of the doors open onto corridors. A small number of them open onto rooms the teacher did not know were in the building."

James Okafor had been Professor Emeritus for five years and was still in his office at eight-fourteen on a Friday night because the retirement had not taken. The department had given him an office to clear out and he had not cleared it out and the department had, after eighteen months of politely not asking, stopped asking. The office was on the fourth floor of the Cognitive Sciences building at Rawlings University, in a corner that got the late light through two windows, and James had been working at the same desk in that corner since 1994.

The email came in at eleven-forty-three in the morning, California time, which was two-forty-three in the afternoon on the East Coast. James had been teaching a graduate seminar on the history of connectionism until one-fifteen. He had gone to the cafeteria. He had eaten a bowl of the lentil soup the cafeteria made on Fridays, because the woman who ran the soup station knew he liked it and because eating what a person who knew you liked had made for you was a courtesy James had learned, from his mother, to extend to any person who extended it to him first. He had come back to the office at two-fifty. He had opened his email at two-fifty-two. The email from Maya Voss had been third from the top, below a circular from the dean and a reminder about a library late fee.

He had read the subject line. Then he read it again.

Subject: About your 2003 paper — I think someone stole it

He had not clicked on the email for almost a minute. He had sat at his desk with his hands on the desk and looked at the subject line on the screen and done the thing he had been doing, at his desk, for most of his professional life when a piece of information arrived that was not what he had been expecting, which was to wait for the first reaction to pass and the second reaction to arrive, because the first reaction was almost always wrong and the second reaction was almost always where the thinking lived. The first reaction, in this case, had been a small stupid flare of something that was not quite hope and not quite vanity but had something in common with both, because his 2003 paper had been cited fourteen times in sixteen years and the fourteen citations had all been from people who had not read it properly, and the subject line in front of him contained the phrase I think someone stole it, which was the first evidence he had seen in sixteen years that a human being had read the paper properly.

The second reaction had arrived about forty seconds after the first. The second reaction was fear. The fear had an object, and the object was the word stole, because the implication of stole was that the thing described in the 2003 paper existed in the world, and the thing described in the 2003 paper was not the kind of thing one wanted to exist in the world.

He had clicked on the email.

He had read it through once. He had read it a second time, more slowly, and had paid attention on the second read to the technical claims rather than the shape of the sentences. He had read it a third time and had paid attention to the shape of the sentences, because the shape of the sentences was telling him something about the person who had written them, and the something was that the person who had written them was not a crank. Cranks did not edit. The writer of this email had edited. The edits were visible in the email the way edits were visible in any piece of writing where a writer had cut her own work: the sentences were shorter than the writer's natural sentences would have been, and there was a small seam between the third and fourth sentences where the opening of the email had originally been longer and had been chopped off. James had been reading his students' writing for twenty-eight years. He knew the shape of a cut.

He had read the email a fourth time and had, on the fourth read, noticed the word Maya had used in the technical description, which was dimensional range you characterized as below-threshold. The word below-threshold was the word he had used in the 2003 paper's methodology section, in one sentence, on page nineteen, in a context where the word had been a quiet and deliberate choice on his part because he had not wanted to call it what he had wanted to call it, which was the place where the thing hides. He had written below-threshold and had moved on, and nobody had ever asked him about the word in sixteen years, and the specific word in the specific dimensional range was the kind of thing a person would notice only if the person had been looking at the thing the paper described in a real embedding space, not in a theoretical one.

Maya Voss had looked at the thing in a real embedding space.

James had read the email a fifth time, just to be sure of the fourth reading, and then he had gotten up from the desk and walked to the window and looked out at the campus, which was doing the thing late-February campuses in New England did at three in the afternoon, which was pretending to be a different season than the one it was in. There was snow on the quad. A graduate student was walking across it with a stack of books and the exact posture of a person carrying a stack of books across snow. James had watched the graduate student until the graduate student disappeared behind the science library, and then he had gone back to the desk, and he had not written a reply, because the reply he was going to write was the kind of reply he was going to need to think about for several hours before writing it.

He had taught the rest of the afternoon's office hours. Three students had come by. He had given them the version of himself he always gave students in office hours, which was deliberate and slow and willing to ask follow-up questions, and not one of the three students had noticed that any part of him was elsewhere, because being elsewhere while appearing to be here was a thing James had learned to do from his father, who had been a Methodist pastor in Enugu and had developed the skill in the service of a congregation that required it.

At six he had gone back to the cafeteria for dinner and had eaten without tasting much of what he ate, which was unlike him. At seven-ten he had come back to the office. At seven-thirty he had taken the 2003 paper out of the filing cabinet behind the desk, the physical paper, the offprint copy he had kept because the journal had gone under in 2008 and the PDF on his personal drive was the kind of PDF that had been migrated across five computers and had acquired, along the way, the subtle unreliability of a document that had been carried too far from its origin.

He had read his own paper.

He had not read the paper in full in almost a decade. The reading was an uncomfortable experience in the way rereading any old paper was uncomfortable, which was the discomfort of watching a younger version of oneself make choices one would now make differently, but the discomfort was overlaid, this time, with the new discomfort of reading a document he now knew had been read by exactly one person properly, and the one person was a machine-learning engineer in San Francisco whose name he had not known when he woke up this morning.

He had finished the paper at eight-oh-two. He had sat with it for twelve minutes. It was now eight-fourteen.

He opened a new email.

Everett Marsh had been twenty-four years old in the fall of 1962 and had been, in the hierarchy of the MIT computer science community at that time, a person nobody had yet decided about. He was a second-year graduate student. He had come to MIT from Princeton, and before Princeton he had come from a boys' school in Sussex that had given him a particular accent and a particular set of vowels and a particular habit of ending sentences with a small downward inflection that American academics found either charming or irritating depending on the academic. He had been in the Tech Square building on a Thursday evening in October, in a room that contained a PDP-1 and three other graduate students and a professor named Dr. Henrietta Lindqvist, who had been, at that time, one of the four women on the MIT computer science faculty.

The subject of the evening was not the PDP-1. The subject of the evening was a question Dr. Lindqvist had raised at the end of her afternoon seminar, and the question had been the kind of question that ended seminars early because the students could not stop thinking about it and could not be useful to any subsequent topic.

The question had been: If a machine could be built that processed language in a way that approximated how language actually operates in a human mind, what would stop such a machine from being used to move the mind rather than to read it?

The question had been rhetorical. Dr. Lindqvist had not intended to start an argument. The argument had started anyway, because one of the graduate students in the room was a man from Caltech named Richard Aldman who was, at twenty-six, already convinced of his own rightness about most things and was saying, with the energy of a man who had walked into a party halfway through and wanted to change the subject, that the question was unserious.

"The word move is doing all the work," Aldman was saying. "You mean influence, I assume. Every piece of language influences every reader. A pamphlet influences. A sermon influences. Propaganda influences. None of it moves anyone in the sense you're implying, because move implies a mechanism, and there is no mechanism."

Everett had been sitting on the radiator by the window with his hands in his jacket pockets, because the radiator was the warmest spot in the lab and Everett, who had been in Cambridge for fourteen months, still had not forgiven the climate. He was not the kind of graduate student who jumped into arguments. He had spent most of his first year at MIT listening, on the theory that listening was the part of graduate school that taught you which of the people around you were going to matter, and that jumping into arguments before you knew who was going to matter was the kind of mistake that was hard to walk back.

"There is a mechanism," Everett said.

The others turned toward him. Dr. Lindqvist, at the front, put down the chalk she had been holding. Aldman, who had been mid-sentence, stopped being mid-sentence.

"Say more," Dr. Lindqvist said.

Everett took his hands out of his pockets. He did not get up from the radiator, because getting up would have changed the geometry of the conversation and he did not want to change the geometry of the conversation. He wanted to say what he was about to say from exactly where he was sitting, because what he was about to say was the kind of thing he had been carrying around in his head since he was nineteen, and the saying of it was going to change something for him whether or not it changed anything for anyone else listening.

"Language doesn't only carry information," he said. "It carries pattern. Rhythm. Cadence. The rhythm is below the information. A reader absorbs the rhythm whether or not the reader is aware of the information. If you could construct a rhythm with specific properties, and you could put the rhythm inside a piece of ordinary language where the rhythm would be invisible to a reader paying attention to the information, the rhythm would reach the reader regardless. What the rhythm did to the reader would depend on the rhythm. And you would not need the reader to consent to any of it, because consent is a function of attention, and attention, in this case, would be aimed elsewhere."

He stopped. The lab was quiet in a way it had not been a minute ago.

Aldman opened his mouth and then closed it. Dr. Lindqvist was looking at Everett with an expression Everett could not read but would remember for the rest of his life, because the expression was the first time an adult in an academic context had looked at him as if he had said something she needed to think carefully about before she responded.

"Mr. Marsh," she said. "Where did you get this idea."

"I thought of it," Everett said. The answer was true and was, in the saying of it, also slightly a performance, because Everett had been raised in a school system in which the phrase I thought of it was a phrase that a boy was supposed to say with a small deliberate modesty, and the deliberate modesty was part of the truth and was also a way of hiding part of the truth, which was that Everett had thought of it while reading, over the previous summer, a book by a Swiss physician named Gerrit Klaes that had been published in 1692 and that Everett had found in his grandfather's library in Sussex, in a locked cabinet that Everett had unlocked himself one rainy afternoon in August because his grandfather had been in London and the cabinet had been the only thing in the house Everett had not yet opened.

He was not going to mention the book. He had not mentioned the book to anyone at MIT and was not going to mention it, in any form, to anyone.

Dr. Lindqvist was still looking at him. The other two graduate students, one of whom was a man named Peter Chen and the other of whom was a woman named Marjorie Vance, were also looking at him. Marjorie was twenty-three years old and was looking at Everett Marsh with the expression of a young woman who had just heard a man her own age say something she wished she had said first.

"Mr. Marsh," Dr. Lindqvist said again. "If what you have just described is possible, do you think it is a line of research a person should pursue?"

Everett thought about the question for what was, by the standards of the evening, a long time. He thought about his grandfather's library and about the locked cabinet and about the book in the cabinet and about the unease he had felt, in August, while reading the book, which was the unease of a nineteen-year-old encountering, for the first time, a piece of writing that was smarter than he was about a subject he had not known could be written about.

"No," he said.

"Why not," Dr. Lindqvist said.

"Because the person doing the research would not be able to control who later used the research. And the category of people who would later use the research is a category I would not trust with a hammer."

Aldman snorted. Dr. Lindqvist did not look at Aldman. She was still looking at Everett, and the expression on her face had shifted a small degree, and the shift was the shift of a professor deciding which of her students she was going to invest the rest of the semester in and which of them she was going to let drift.

"Come to my office on Monday at two," Dr. Lindqvist said. "Bring the thinking. I want to see where it goes."

Everett went to her office on Monday at two. He went on Monday at two for the rest of the semester, and then he went on Monday at two for the rest of his second year, and then he went on Monday at two for the rest of his third year, and then he did not go on Monday at two because Dr. Lindqvist had died in her sleep in July of 1965 of an aneurysm nobody had known she was carrying, and Everett, at her funeral, had been unable to speak when it was his turn to speak, and had been the only one of her students who could not get through the tribute he had prepared. He had finished the Ph.D. in 1967. He had been offered a position at MIT and had turned it down, because he had decided, in the two years after Dr. Lindqvist's death, that the work he wanted to do was not the work that universities funded in 1967. He had gone to work for a private research firm in New Jersey for a few years, and then for a firm in Massachusetts, and then for a firm in California, and by 1978 he was in his early forties and was living in a house in Menlo Park with a wife and a twenty-year-old son named Gregory, already in his second year at Berkeley.

None of that mattered, on the Thursday evening in October of 1962 in the room with the PDP-1. What mattered, on that evening, was that Dr. Lindqvist had said Come to my office on Monday at two, and that Everett Marsh, at twenty-four, had just made a decision about the rest of his life, and that the decision had been the decision to not do the thing he had just described. The not-doing was harder than the doing would have been. It required him to know how to do the thing and to know it well enough to recognize when anyone else in the world was doing it, and the recognizing was a thing he carried in his head, alone, and would keep carrying.

He carried it into the Geneva library in 1996 and showed it, in the form of a single sentence he had been rehearsing for seventeen years, to an eleven-year-old boy named Declan who had been sitting across from him at a reading table under a green-shaded lamp. The sentence had been: The interior is the part of a person that is easiest to move without the person knowing she has been moved, and people who study the interior for a living are either doctors or something else, and you want to know which one you are dealing with before you let them near you. Declan, at eleven, had filed the sentence without understanding it. Everett had not expected him to understand it. Everett had expected him to file it, which was the thing Declan had been raised, from the age of six, to do with anything Everett said in the Geneva library, and the filing had been the entire point.

Dr. James Okafor did not know any of this on the evening of the Friday he received Maya Voss's email. He did not know about Everett Marsh or Dr. Lindqvist or the PDP-1 room in Tech Square. He did not know that Maya's mother had once worked at Calyx Systems. He did not know that Everett's one visit to Calyx Systems in 1979 had been a visit in which Everett had looked at a particular piece of software the Calyx engineers were working on and had said nothing during the visit and had, on the flight home to Menlo Park, written, in the small black notebook he carried in his breast pocket, the sentence They have built the thing. May God have mercy on all of us.

James did not know any of that. What James knew, at eight-fourteen on a Friday night in his office at Rawlings, was that a woman in San Francisco had written him an email at two-forty-three in the afternoon that had told him, without telling him, that the thing he had reasoned about theoretically in 2003 had been built, and that it was running, and that it was inside a production machine-learning pipeline at a company in California, and that the woman who had written the email was the person who had found it, and that the person who had found it had done the thing James would have hoped the person who found it would do, which was to narrow her search carefully and to contact the one human being in the world whose paper described what she was looking at.

He typed a reply.

He did not type a long reply. Long replies, in his experience, were the replies of people who had not yet decided what they were going to do. He had decided what he was going to do, and the reply was going to be the reply of a man who had decided.

Subject: Re: About your 2003 paper — I think someone stole it Ms. Voss, You are not wrong. I am going to ask you several questions. I am going to ask them one at a time, over the next several days, and I am going to ask you to answer them on a phone I will give you the number for, from a location you have not used for any of the work you have already described. I will pay for the phone if that is necessary. Please confirm that you are willing to do this before I send the first question. Appendix B exists. I have a copy. It has never been digitized. It will not be digitized over email. If we are going to have this conversation about what is in Appendix B we are going to have it in person, and I will come to you, because you should not travel on my account. One more thing. You said in your email that if you are wrong you will apologize for wasting my time. You are not wrong, and even if you were wrong, you would not have been wasting my time, because a person who has read my paper closely enough to notice what you noticed is a person I have been waiting for for sixteen years. I did not know I was waiting. I am writing this from the desk where I wrote the paper. I am writing it at eight-eighteen on a Friday night. I want you to know, before we go any further, that I am glad you wrote to me. James Okafor

He read the reply once. He changed one word. He read it again. He sent it.

He sat at the desk for another minute after sending it. The snow on the quad had gone blue in the way snow went blue at eight-twenty on a winter evening in New England. The office was quiet in the way his office was always quiet on Friday nights, which was the quiet of a building emptying around him, and the building emptying around him had been, for twenty-eight years, one of the things he liked best about Fridays. Tonight the emptying felt different. Tonight the emptying felt like a thing that was going to matter, because the emptying was the sound of a building in which nobody was listening.

He got up from the desk. He put the 2003 paper back in the filing cabinet behind him. He turned off the desk lamp. He stood at the window one more time and looked at the quad, and then he picked up his coat and went home.

He was sixty-two years old. He was going home to an apartment where nobody was waiting for him, and the nobody-waiting was, for the first time in several years, the exact thing he needed. The work he was about to do required a kind of solitude he had not had a use for in a long time.

He locked the office door behind him. The building was quiet. Down the hall, the light was still on in a graduate student's carrel, because graduate students' carrels had lights on them at all hours on all nights, and the light in the carrel was the ambient evidence of the category of work James had chosen for himself at twenty-eight and had never walked away from, and he walked past the light and down the stairs and out into the snow.

The air outside was cold in the way Rawlings was cold in late February. James put his hands in his coat pockets and walked to his car, and on the way to the car he thought, for the first time in sixteen years, about the room in Tech Square at MIT in 1962 where his own advisor, Dr. Lindqvist's last surviving student, a man named Peter Chen, had told him, in 1989, about an evening in October of 1962 when a graduate student named Everett Marsh had said something on a radiator by a window that had changed the way Peter Chen thought about what a machine could do to a person. James had, in 1989, measured the story against the rest of what he knew about the early days of the field and set it down, and had not thought about it again, because the story had been one story among many.

He got into the car. He sat with his hands on the wheel for a long time before he started the engine.

[End of Chapter 4]