Brad Segro and the Great Information Virus.
In this three-part history, I propose a re-writing of how we understand the decline of our predecessors based on the remarkable findings within the recently excavated journals of Brad Segro, the progenitor of the virus that wiped out information technology in the third century Before Descendance. Selections of the thirteen volumes discovered pertain to Brad’s involvement in The Event, as it is commonly dubbed, that so dramatically changed the course of human history. Much of that information has been distilled into a coherent narrative in the first section. The second will consider the difficulties of working with Brad’s journals, and begin to juxtapose what they don’t tell us with what they do. The third and final will finish discussing the journals’ limits, and in their context propose questions to guide our research moving forward.
At the turn of the last century, there was an undergrad student named Brad Segro, who studied literature and computer science, among other things. Brad read a lot of interesting articles about so-called “deep learning” computer programs that were tackling problems of natural language processing—predicting an author based on a sample of text, for example, or translating from one written language to another. There were even some programs that could generate text or images after reading a great deal of examples.
Brad was excited by the possibilities of this technology, but he was always disappointed to read the texts produced. He saw them as poor imitations of the human art of writing and came to believe the machines generating them were far from capable of telling meaningful stories. He did a little reading and decided that since machine learning algorithms generally improve with larger samples, he could address the problem of meaningful creation by building an artificial neural network that would read every single text, ever.
That’s not exactly what he accomplished, of course. He exclusively considered texts presented in language, ignoring texts with other forms such as images or audio. Brad also limited himself to such texts as available on the internet; he figured there was enough material uploaded already and recognized that whatever small number texts he might upload individually would fail to have an impact in such a large pool. He did not exclude any languages, however, resulting in a sample of over 40 languages majorly represented.
Of course, with billions of pages and exabytes of data, Brad’s learning program toiled for weeks. In the interim, Brad’s term of studies ended. He left his computer running and went outside for a while. When he came back one day, he was surprised to find that his program had output some text. The following symbols had been printed to Brad’s terminal:
Brad looked it over briefly. He was not a linguist. Even if he was, this small output would not be enough to derive meaning from the symbols. Brad adjusted the parameters and instructed the program to output some more, an amount that would have printed to nearly 100 pages. He wrote a little note explaining what it was and sent it to his university’s linguistics department.
At this point, the information in Brad’s journals begins to intersect with what we’ve compiled from our predecessors’ public records. Interestingly, little of what Brad documented during this period actually concerns his machine. He took quite a lot of notes—40 pages in the week following the completion of his program—but the vast majority of those concern a romantic relationship he was having at the time. When he does mention his machine, it is very much within the emotional modality in which he had been writing—he expresses curiosity as to its products, anxiety that they won’t be of value, and a rich pride engendered by the faint hope that his program would be a success. He speculates wildly as to the fame and wealth that it would bring him, even while acknowledging to himself that it was too early to tell.
What Brad did mention is that he got a little note back from the department chair right away, saying, “Thanks for the email,” and that they would take a look at it. On the same day that the Springfield Chronicler reported that the bodies of the missing linguistics department members had been found, Brad writes the final entry in the last volume of the journals:
Did he really not think that dude was flirting with him? Is he really that naive? Or is he just defensive because of the way I accused him?
“Entre las formas que van hacia la sierpe
dejaré crecer mis cabellos”
“The wackness is spreading like a plague”
The scents are maddeningly bland. I pull air from my hand and smell the sweat Its musk grows and I veer towards sanity
“n k’aba’a a’an Yack. xik kue sa’ kuakax nin chal tz’ibitz patux”, right? I return to her letters and it’s all unraveled.
if only… It isn’t nearly real enough. It’s not that simple. But maybe I will get out of town for a while”
It’s tempting to try to read some sort of meaning into these messages that could connect them to his program, which was beginning its catastrophic work even as he was writing this, but the reality is that they don’t tell us much at all. Reading this single page is representative of what it was like working with his journals as a whole—parsing masses of irrelevant personal content for slight clues, trying to calculate the truth from the probabilities of many unreliable statements. It’s possible that Brad did get out of town, which might explain why he suddenly stopped journaling, but there are many other plausible reasons for this phenomenon, and none can be ruled out without further evidence. It was only a few days later that the last issue of the Springfield Chronicler was printed, and other media organizations began reporting the destruction of Brad’s university, followed by many others. Brad does not appear again in any of these records, so we have no idea what happened to him after this date. Nor do we know the extent to which he was affected by his own creation, or if he or any of his contemporaries ever discovered his role.
If we imagine the spread of the virus across a network of institutions responsible for the production of knowledge, then most of what we know about it comes from nodes in the network just outside of the portion of the network already affected. Unfortunately, some elements of the virus spread faster than the system’s own ability to recognize it—for, as we all know, institutions of storytelling showed signs of the madness even as they reported on other institutions suffering from it. What’s so frustrating about Brad’s journals is that they’re so close to the absolute center, and might be the first documents we’ve discovered related to the virus that were unaffected by it, but they don’t show so much as an awareness that the virus existed. Their position in relation to our historical problem is a great tease, and although what we have been able to draw from them about his program is critically illuminating, they are ultimately a disappointment.
The monumental significance of the discovery of Brad’s notebooks is matched only by the challenge of actually making sense of it. Most importantly, we now know for sure that The Event was caused by the work of a single human being. Folklore and popular fantasy will no doubt draw great meaning from this realization. But I am a historian—we are historians—and we must resist the temptation of speculation and sensationalism. Brad’s importance is that he documented his life in a manner that survived the Great Information Destruction, and perhaps more miraculously, that the paper on which he wrote survived for over three hundred years.
Not long ago, human society reached a level of informational complexity that is daunting to conceive of. The proliferation of computers and the “Internet” that connected them enabled a vast production and sharing of knowledge, which at its onset was theoretically democratic in that almost any individual with a computer could produce information and share it with anybody. In the span of a few years, however, the sheer volume of information produced and shared made it impossible for the consumer of information to navigate it in any meaningful way. Thus arose institutions of several classes to assist the consumer in finding relevant information: one was the search engine, which directed the consumer to sites containing information it deemed useful based on a query by the user; another was the sites themselves, often maintained by the same groups of people that had controlled the distribution of printed knowledge before the computer era. These two classes worked in tandem to empower certain information and disempower the rest by its placement in highly trafficked sites.
It’s interesting to consider what we can glean about Brad’s program from his journals in the context of the informational structure he inhabited. His program read everything, including the information his society had sought to marginalize and effectively destroy. Perhaps humanity was not ready for the iterative power of the machine to deliver unto it everything it had tried to throw away. This would have fascinating implications for the nature of the relationship between dominant and oppressed literatures, implications I’m sure will be seized upon by socio-information theorists. I caution against this for now, for we can’t confirm these ideas until we better understand the mechanism by which the virus worked.
Oral tradition tells us that our ancestors destroyed everything the victims of the virus ever wrote, or even touched—we thought we’d never see the symbols until we found Brad’s journals. As far as we can tell, the small printout he kept wasn’t enough to affect us, and we assume that it didn’t affect him, so hopefully our institutions of knowledge will still be around when the next piece of the puzzle is unearthed. If not, I apologize for sharing this paper with you. Let us hope that our predecessors fully paid their debts, and that we may survive to continue the pursuit of our past and of the truth.