Fall; or, Dodge in Hell

by Neal Stephenson

  She shrugged. “Okay, should we talk about that?”

  “Yes. This is your so-called capstone project. Or keystone. Whatever you want to call it.” Solly glanced out the window of his office for a moment, trying to access the relevant memory from the tiny compartment of his brain where he kept track of administrative minutiae.

  “I believe ‘senior thesis’ is the term they’ll be looking for,” Sophia ventured, since Solly appeared to be looking in the direction of the dean’s office.

  “Very well. You incubated it last summer, during your internship in Seattle.” Solly glanced at the screen, eliciting a nod of confirmation from C-plus.

  “That’s a nice way of saying it. What really happened was that I got in way over my head there. I had no idea what a big deal it was, how long it would take. I barely even got started before I had to come back here.”

  “You incubated it last summer,” Solly repeated, raising his hands to make air quotes, and glancing again toward the dean’s office. “Building on the foundation of knowledge, and the code base, that you established during those months, you returned to Princeton in the fall with a clear understanding of how to move forward.” Again he looked at the screen. C-plus and Enoch were both poker-faced. No objections.

  “Yes. Absolutely. Whatever.”

  Solly nodded toward the pad of paper that was sitting on the table in front of Sophia and paused for a few moments while she jotted this down. He’d been speaking toward the screen, perhaps worried his voice wasn’t coming through, but now dropped into his usual relaxed tone. “I have looked through DB myself a few times, and I know it’s an Augean stables.”

  “Aww, thank you, Solly!”

  “Huh?” C-plus asked.

  “We D’Aulaires’ fans have to stick together,” Solly explained, and winked at Sophia. For the two of them had long ago bonded over their shared love of those books. “Anyway. Dodge’s Brain. My god. All of those different data formats at war with each other. Each of them encoding someone’s pet theory as to how the brain works. It’s no surprise you spent the whole summer just getting oriented. The only thing that matters is that the Forthrast Family Foundation didn’t give you the sack. You came back. You were able to do some new things here, in this department. Now, tell me in your own words what you did, and we’ll bang it into a form that is acceptable as a senior thesis. Which”—and he checked his watch—“is due—”

  “In a week. I know.” Sophia threw her head back in embarrassment. “I just don’t know where to begin.”

  “‘Dear Mom,’” suggested Solly.

  “Huh?”

  “That’s how to begin.” Solly made a sort of exaggerated pointing gesture, apparently trying to draw her attention to the pad of paper. “‘Dear Mom.’ Write that down.”

  She wrote it down.

  “‘I’ve been working on digitally simulating my great-uncle’s brain,’” Solly continued.

  “I’ve been working . . . ,” she muttered, writing as fast as she could. “. . . brain.” Then she looked up expectantly.

  Solly shrugged. “I have no idea what to say next. You know your mother, and your project, better than I do.”

  She wrote a sentence.

  “Now we’re cooking,” C-plus said. “I like this plan. Read it back to us.”

  “‘When I got back to Princeton in September, thanks to all the help from C-plus, I feel like I knew my way around DB pretty well,’” she read.

  “Flattery will get you everywhere,” Corvallis said.

  “Tell you what,” Solly said, manipulating some kind of UI. “Why don’t you dictate it. Saves time. I just turned on voice capture.”

  “The connectome was a Tower of Babel,” Sophia began, speaking slowly at first, then warming to the task. “The same basic set of connections, interpreted and expressed in dozens of different ways. In order to even get started, I had to write code that would walk through all of those files and spit out a connectome that I could at least work with. During the summer I’d laid a lot of the groundwork, but the code was slow and buggy. I cleaned it up and got it to run faster. By Thanksgiving break I had something I could use.”

  “Use how?” Solly asked.

  “I just mean that it was compatible with the neural simulation algorithms that I had access to here at Princeton. Until then it had been a total ‘square peg, round hole’ problem.”

  “But by Thanksgiving you were in a position to get the peg into the hole.”

  “Yeah, so then I fired it up on a small scale during December. Meaning, I got the simulation algorithms to run on a tiny subset of the connectome.”

  “Proof of concept by Christmas. Very good.”

  “Okay, that’s what I’ll call it. Proof of concept.” She jotted that down. “Then I spent winter break feeling kind of depressed about the whole thing.”

  “No one gives a shit about your feelings,” Enoch pointed out, in a cheerful way. Not being mean at all. Wry humor.

  Sophia accepted it in the same spirit but pushed back a little. “Right, but there was a technical basis for those personal feelings.”

  “Which was?” C-plus asked.

  “I couldn’t think of any intermediate steps to take next. It makes sense to simulate one neuron. Two neurons talking to each other. Fine. Beyond that, it’s a network. Network effects are all that matter. Simulating, say, a thousand neurons doesn’t break any new ground. People have been doing that for decades. The fact that the connectome of those thousand neurons was arbitrarily cut and pasted from DB is completely meaningless—it might as well have come from a mouse brain.”

  “Agreed,” Solly said.

  “The only meaningful next step was to light up the whole connectome at once.”

  “Really? Then why had no one done this before?”

  “Well, for one thing, the data has been lacking. A full connectome of a human brain has only come into existence in the last couple of years. Then there’s the lack of resources—who would pay for all the computing power required? So it was inconceivable until Hole in the Wall came online.”

  “That’s good,” Solly said. “Maybe you are being diplomatic by leaving out another factor, which is the academic mind-set. Academic science advances in many micro-steps, one paper at a time. Peer-reviewed papers are the way we keep score. The more papers, the better. If you can take a project and slice it fine, like prosciutto, you publish more papers, and your score goes up. But you were behaving like someone who didn’t know or care about that.”

  Sophia blushed. “Maybe I’m naive. Okay.”

  “You’re rich,” Enoch announced.

  “I beg your pardon?”

  “You’re rich,” he repeated. “There’s nothing wrong with it. It’s not a criticism.”

  “How does that enter into it?”

  “Two ways. First, you simply don’t care about playing the game—publishing as many papers as possible,” Enoch explained, with a nod at Solly. “Second, you had access to the resources needed.”

  There was a long silence, during which it seemed entirely possible that Sophia might start crying.

  “Listen,” Enoch said, “there is a long and honorable history, dating back to the Royal Society, of the gentleman scientist. And now the lady scientist. We don’t like to acknowledge it because we wish to maintain a polite facade of egalitarianism. But there’s a reason why so many important theorems are named after members of the titled nobility of Europe.”

  Sophia didn’t answer, but she seemed to be settling down a bit.

  “You don’t have to put Enoch’s observations into your thesis, of course,” Solly said. “But you should be prepared for people to ask the question.”

  “If that happens,” she said, “I may not have satisfactory answers.”

  “How so?”

  “I still don’t really know where the money came from.” She said it loudly, with a sidelong glance at the screen.

  “An anonymous donation,” Enoch said. “Happens all the time.”
<
br />   “Is that how I should say it? In my thesis?”

  “Be as brief as possible, Sophia, I beg of you,” Solly said. “It gives me a headache, trying to sort out the Forthrasts and the Waterhouses and the Shepherds and all of their interlocking bits. It’s like binge-watching a Mexican soap opera.”

  “Well,” Sophia said, “between you and me . . .”

  Solly shut off the recording.

  “Over Christmas I spent a little time with my uncle Jake and his family. He asked me what I was up to and I told him. By the time I got back to Princeton after the break . . .”

  “You had an account at Hole in the Wall,” Solly said.

  “Yes.”

  “Set up in your name by a mysterious benefactor.”

  “Yes. And when I logged on to it, I was able to see the available balance—the amount of money I had to spend. It was a pretty impressive number.” She looked at Enoch’s face on the screen, but he seemed to be preoccupied by something in the charming pub—perhaps an engaging barmaid, perhaps a soccer game on the television.

  “We’ll clean this up verbally,” Solly mentioned.

  “I beg your pardon?”

  “I’ll help you tidy up the language so that it better befits a Princeton senior thesis.”

  “Right. Thank you.” She knew what he really meant: she’d better tidy it up herself, then run it by him. “Anyway, there was enough money in the account that I was able to transfer all of the files from our servers here, where they’d been living, to local storage at Hole in the Wall.”

  “Much faster,” C-plus translated, “a more solid platform on which to base further work.”

  “Yes,” Sophia agreed, jotting this down. “Then there was a couple of months of tweaking the code and the algorithms to run on their devices.”

  “We originally wrote the code to run on conventional computers,” Solly said, translating this. “Quantum computers require different strategies, different optimization.”

  “That’s very much a work in progress, by the way.”

  “Of course. Fuck’s sake, Sophia, it’s just a senior thesis. A simple but workable port is all any sane person would expect.”

  “Well, that’s what I did. The bare minimum that would run. With a lot of help from your postdocs.”

  “Who will of course be copiously acknowledged.”

  “That’s the only part of the thesis that’s actually written!” she replied, beaming. “The acks.”

  “So, now we are up to, what—?”

  “Mid-April.”

  “About a month ago. You got it running.”

  “I got a neural simulation running on the Hole in the Wall system with one neuron. Then two, with the neurons passing messages back and forth.”

  “With each being a separate, independent process—that’s important to mention.”

  “Right.”

  “And everything bolted down nice and proper—the messages all being transmitted through modern cloud protocols, suitably encrypted and verified,” C-plus reminded her.

  “Of course. So that it would scale.”

  “Very good. And then?”

  “Then I scaled it.”

  “You flipped the switch.”

  “Yeah.”

  “The big red-handled blade switch on the wall of the mad-scientist lab.”

  Getting into the spirit, she said, “Yeah. I hooked the jumper cables up to the neck bolts.”

  “You turned on the whole DB at once,” Solly said, “and the lights dimmed in Ephrata, or something.”

  “Maybe. I’m not so sure about the lights dimming.”

  “What happened then?”

  “Little wiggles.”

  “Hmm. I’m afraid you’ll have to do better, Sophia.”

  “Okay . . . an oscillatory phenomenon?”

  “What precisely was oscillating?” C-plus asked.

  “The burn rate.”

  “That is a Silicon Valley term that might not go over well in a senior thesis, young lady,” Solly said, mock-serious. But serious.

  She looked sheepish and drew breath a few times as if about to speak. Each time she thought better of it.

  “‘Dear Mom,’” Enoch prompted her, and took a swig of beer.

  “Yes,” she said. “See, Mom, the thing is, it’s one thing to simulate a brain. It’s another thing to talk to it. We forget this because our brains are hooked up to bodies with handy peripherals like tongues that can speak and fingers that can type. As long as those work, you can always get some idea of what’s going on in a brain. But a brain in a box doesn’t have those. Which is probably another reason that no one has tried this yet.”

  “Namely, that there’s no point in doing it if you can’t get meaningful information out.”

  “Yeah, it’s like hooking an EEG up to a patient in a coma,” Sophia said.

  “That is a useful analogy.”

  “The EEG is a legit scientific instrument. It’s definitely telling you something about brain activity. No question about it. But you can stand there all day looking at the strip of paper coming out of the EEG, staring at the wiggles, and still not have the faintest idea whether anything useful is actually going on inside of that brain. Whether the person is conscious. What they’re thinking.”

  “And here the EEG is an analogy for what exactly?” C-plus asked. He knew, of course. He was prompting her.

  “The burn rate. The cost. The monetary cost of running the simulation. When a lot of neurons are engaged and a lot of signals are being passed over the connectome, the quantum processors at Hole in the Wall are busy, and so the system charges us more money, and draws it down from the account. I can log on and watch the balance in the account declining. Sometimes it declines fast, sometimes slow. If I draw a graph of the burn rate, it looks like this.” She peeled the top sheet off the pad to expose a new page and slashed out a long horizontal axis that presumably represented time. A shorter vertical axis was burn rate. Starting at their origin, she began to draw a line across the page, wiggling the pen randomly up and down. “At first, a lot of what looks like noise. Activity comes and goes in surges. I tried running some statistical packages against it but it’s pretty close to just chaos. But then later we start to see some periodicity emerge.” She began ticking the pen up, down, up, down, in a steady rhythm. “Like a heartbeat. Just for a little while. Then back to random.” More pen shaking. She was approaching the edge of the paper. “Just within the past week, though, it really settled down.” She slowed the pen’s progress and marked out a neat, regular series of ramps, each ramp ending with a sudden drop toward zero. The last of them went off the page. Having run out of paper, she flipped the pad around and held it up so that Enoch and C-plus could see her handiwork.

  “And that’s where we are now,” she concluded. “As of an hour ago.”

  “Like it’s stuck in some kind of repetitive cycle,” C-plus said. “That’s my takeaway from your graph.”

  “That’s what worries me,” she said. “Or maybe it’s waking up, building to some level of activity, and then going to sleep again.”

  “I would not recommend putting that kind of anthropomorphic language into your thesis,” Solly said.

  “Of course.”

  “A very basic question: can the connectome self-modify?” Enoch asked.

  “Self-modification is a vast, separate thing from what I’m trying to do at this stage!” Sophia exclaimed. She glanced at Solly and C-plus and read in their faces that they were both as taken aback as she was by Enoch’s question. Oh, it was a fascinating topic to think about. But way out of scope for this project.

  “‘Dear Mom,’” Enoch repeated.

  “Okay. Okay. Mom, keep in mind there’s a huge difference between this simulation and a real brain. A real brain is a Heraclitean fire. Never the same brain twice. You can take a snapshot of the connectome at the moment of death—which is what we have done here in the case of DB—but that bears the same relationship to a real brain as a freeze-f
rame does to a movie. In a real brain, every single time a neuron fires, the brain rewires itself a little bit in response to that event. Frequently used connections get strengthened. Neglected ones atrophy. Neurons get repurposed. Things get remembered—or forgotten. And none of that is happening in my simulation.”

  “Why not?” Solly asked. He knew the answer perfectly well but was still acting as devil’s advocate. Also, helping to shield her a bit from Enoch’s weird and transgressive line of questioning.

  “Well, first of all, because I didn’t feel like I was ready. I had to get it running first. To see what happened. Whether it would work at all. Secondly, because it’s more expensive—it would consume more memory, more processing power, and I was afraid of running out.”

  “The algorithms you are using have the self-modification capability built in,” Enoch pointed out. “Unless you went to the trouble of ripping it out.”

  “No, I didn’t rip it out. I just turned it off. Suppressed it for now.”

  “I don’t think anyone here is arguing with that decision,” Solly said, with a sort of quizzical glance at Enoch. “You’re right that you had to simply get it running first—to see what would happen. And no one will blame you for being conservative with your expenditure of resources.”

  Enoch said: “But it’s hardly surprising, is it, that you’re seeing repetitive, cyclic behavior. That it’s stuck in some kind of loop.”

  “Freeze-framed,” Sophia said.

  “Can I see it?” Solly asked. “Can you log on for me? I’m just curious to see what this all looks like. You’ve piqued my curiosity.”

  “Absolutely. It’s a pretty simple old-school interface. Can we use that?” She indicated an empty space on the wall of his office and slipped her glasses on.

  “That’s what it’s there for,” Solly said. He got up, went over to his desk, and found a wearable rig half-buried in clutter. By the time he had put it on and booted it up, Sophia had placed a virtual screen on his wall and was logging in to her Hole in the Wall account. This was disconcertingly old-school, looking like a circa-1995 web page, enlivened with a background photo of the eponymous coulee. Meanwhile, in another window, she was booting up the program she’d been using to plot and analyze the burn-rate data. The first thing that came up was a graph, looking generally similar to the one she’d just drawn on the paper, but with more noise and complexity. The sawtooth wave pattern was clearly discernible.