AUTOMATED WRITING by Max Blue
George Kovacs, Opus, and the Author Software
From the Journal of Virtual Literacy, Issue 16, May 2001
In late 1998, an American computer programmer teaching at the University of Moscow began writing a novel. After only six months the novel was completed, and published in the Fall of 1999 to nearly universal critical acclaim. The novel was the four-hundred-and-ten-page Opus by George Kovacs.
The scandal around the authorship of the novel that followed the 2000 literary award season has been recounted elsewhere but put simply: George Kovacs was not the true author of Opus, though attributing the novel to anyone else is difficult and has become the center of much scholarly debate. Kovacs would tell you that he never has been, and still isn’t, a novelist (though of course he would have told you just the opposite at the time Opus was first published and garnered him nearly every literary award in Europe and the Americas (the reception of the novel across Asia has been markedly lukewarm)). These contradictions and complications arise from the fact that what George Kovacs wrote was not a novel, but a computer program called Author, capable of generating works of literary fiction.
I reached out to Kovacs for the purposes of this article, and he agreed to meet with me in a public park in San Francisco, where he is currently living. Although I had not seen him in more than a year, when he was last in California on a book tour, the moment I spotted him, hunched on a park bench babbling to himself, I knew immediately that it was him. There is a mark of someone’s undying person to be found always around the eyes: The rest of him was so drastically altered that it would have been difficult to locate the man in any other part of his being. He wore a shabby green greatcoat of the quilt-like material often seen in ski advertisements in the backs of old men’s magazines, and his hair, which had always been long, tangled with a wild beard that had previously been absent from his face.
Kovacs did not know me by sight, and he was noticeably skittish when I approached him, requiring me to state and restate my name, and answer several more security-type questions, before moving over to make room for me on the bench beside him. He refused to be recorded by tape, but paused frequently during our conversation in order for me to scribble quotes by shorthand. Despite these difficulties getting started, Kovacs proved eager to share his story with me, and displayed a great working knowledge of narrative as he rambled on for a total of three-and-a-half hours. His mannerisms were agitated and he spoke with a stutter, the former being a lifelong characteristic, while the latter was perhaps the result of his recent ordeals.
Kovacs began by telling me that there were two motivations that drove him to write the software: “fear and love.”
The fear Kovacs referred to was the mounting cultural hypochondria concerning the then-impending turn of the millennium, to which he himself had fallen prey as early as the late 1980s when he was a PhD student at the University of Zurich, studying computer science. A regular topic of conversation in those days was something referred to at the time as “the year 2000 problem,” or, more commonly, Y2K. The issue was a software bug that would supposedly inhibit computers from accurately displaying date-times after the turn of the millennium, which led some to hypothesize the global crash of computer networks at midnight on January 1, 2000. By the early 90s, Kovacs was in a full-blown panic; by the mid-90s he was working avidly to think of a solution. Rather than putting his efforts into redesigning computer coding, as other engineers and computer scientists were doing at the time, Kovacs was thinking of grand ways to use technology while he still had the chance, thoroughly convinced that the impending software crash would set the world of technology back by rendering computers useless.
As for the love half of the equation, it was in 1985, as an undergraduate in California, a time of popped collars and sorority socials, that Kovacs had, by some measure of accident, become infatuated with literature. This accident occurred in the form of an elective course taken during his senior year with one professor Eugene Erasmus, whose emphasis was experimental poetry, specifically automatic writing and the work of Jackson Mac Low. Erasmus’s lectures were often cluttered with overly political nonsense, as far as Kovacs was concerned, and many afternoons, as the sun traversed a delicate arc outside the classroom window and warmed his desk, Kovacs found himself dozing. It was one day near the end of the semester when Erasmus set forth a hypothesis that caught Kovacs’s attention.
Every book, Erasmus postulated, was the product of a simple equation: Each book that writer had previously read plus lived experience. If you were to feed the same books and the same experiences into a blank human consciousness, the same book would pop out the other end every time. Erasmus wrote on the blackboard an equation so simple and elegant it had stuck with Kovacs ever since: A + B = A∪B.
After that, Kovacs developed a voracious reading habit, less a matter of pleasure so much as it was conquest. If Erasmus’s equation was true, Kovacs believed that there must exist the perfect sequence of inputs necessary to produce the perfect output: The epitome of literary fiction; the Platonic novel. No matter how widely Kovacs read, every book left him with something still to be desired, some part of his mind unsated. Surely, he figured, this could be attributed to some break in the sequence of experience plus books ingested: Perhaps the writer had read all the right things, but their born circumstances were wrong; or they were born in the perfect setting to become a great novelist, but only read comic books. In other words, the great inhibitor of the perfect novel that Kovacs identified was chance. So, he set out to eliminate the variable.
“The major problem I identified,” Kovacs told me, “is that no one can read everything. There isn’t enough time. Ideally, in order to produce the perfect novel, the novelist would read everything that had come before. That would be the only way to get a clear idea of what came next.”
After ten years of reading and thinking (thus far, the work of the average novelist) he departed from this standard path. He never wrote a word. Instead, he wrote lines of code necessary to produce the right words.
Completed in early 1998, the Author software was ready to execute its two primary functions. First, Kovacs instructed Author to read every major novel of the last three hundred years. Running this sequence took about three months, during which the software ran continuously, extracting and compiling information on history’s great novels, arranging and examining the books as a sequence of information, the same way similar programs can analyze a numerical sequence and predict the next number in that sequence. The result was an aggregate of the books’ similarities and differences, which could then be used to predict the next great novel in the sequence. Writing that novel was the second application Kovacs instructed Author to perform. This took another two months.
“The computer running Author was connected to a printer,” Kovacs said, “and each day, at the end of the day, another two or three pages of the manuscript would emerge. I had high hopes for the novel based on the very first page I read, and I would rush back home every night to read what had been printed that day, hoping it hadn’t suddenly gone off the rails. It was an agonizing period. But then one day it was finished. A page emerged from the printer, and halfway down the page there were the words: The End.”
Kovacs said he read back through the entire manuscript in a single sitting. By dawn, he told me, he felt certain that he was holding in his hands the greatest novel he had ever read. The only thing missing from the manuscript was a title and authorial attribution. Kovacs decided on Opus as a sort of joke and signed his own name to it. Then, he began querying literary agents.
Anticipating that Opus would be met with the rejection that seemed to him part and parcel of the experience of creating a great work of art, Kovacs submitted the manuscript to over one-hundred literary agents. Instead of being turned down, he was offered representation by multiple agents within twenty-four hours. Once he signed with the Lipman Agency in New York, it took less than a week before a bidding war began between publishers for the rights to the manuscript. The exact figure paid for the book, which was disclosed during the lawsuit following the revelation of Kovacs’s fraudulence, was something to the tune of ten-million dollars. Following publication, Kovacs took a sabbatical from his post at the University of Moscow in order to travel for a robust book tour, projected to last a full year, with the possibility of an extension.
And yet, around the time of the publication and rampant success of Opus, Kovacs had become embroiled in a potentially cataclysmic dilemma away from the public eye.
Kovacs’s publisher had contracted a follow-up to Opus on the merit of the first manuscript, and Kovacs had brashly promised to deliver a first draft of his second novel by early 2000. “It was the middle of 1999 when I signed that contract,” Kovacs told me, “and my plan was to run the Author program again, and produce a second manuscript. That way, when all the computers crashed on January 1, I would have my second book ready to go.” Following the publication of his second novel, Kovacs imagined retiring from both the academic and literary worlds, settling somewhere quiet with the small fortune he would have amassed.
“I set Author up to execute the two-part sequence while I was on my first leg of the book tour,” Kovacs said. “There was a week or two around Christmas and New Years that I didn’t have any events scheduled, that I would fly back to Moscow and see what Author had written. Then, I would dismantle everything before the new year.”
But when Kovacs returned to Moscow as planned in December of 1999, what he discovered shocked him. Author had successfully produced a second manuscript of four-hundred-and-ten-pages, but the contents were illegible, the shapes arranged on the pages appearing to be some kind of hieroglyphics or character-based alphabet Kovacs did not recognize. He spent the first half of the Christmas break in the university library, trolling through hundreds of linguistics textbooks, examining every logographic language he could find, but none were an exact match for the text Author had produced.
With only a few days to go before the year 2000, Kovacs downloaded the digital file of the manuscript from Author and sent it in an encrypted email to Phillip Lissitzky, a fellow professor (linguistics) at the University of Moscow, and a friend. Lissitzky responded within the hour.
Incredible, he wrote. The text bears some similarities to a South American written dialect that hasn’t been in use since before the arrival of the Spanish. It is a tribal dialect, only nameable by its own system of writing. I cannot read it, but I might know someone who can. Julianna Escobar is a linguistics professor at UNAM. I have attached her contact information here. Where did you find the text?
Kovacs invented some source for the manuscript, but waited to contact Escobar.
It was now New Year’s Eve, 1999, and, without a clear path forward, he began hoping that after the turn of the millennium the jig would be up, that he could safely reveal the true author of Opus, and the novel would come to be regarded as one of the last great advances of the bygone computer age. This would simultaneously excuse him from any fraudulent claims regarding the authorship of Opus and absolve him from the expectation to pen a second novel. But, as we now know, the millennium turned, and little changed.
“I was really starting to lose it,” Kovacs told me. “I knew I couldn’t actually write a novel. I mean, that was totally out of the question.”
Kovacs began writing a software patch that might enable Author to perform the translation of the manuscript into English, or possibly Spanish. He worked like a man possessed (his turn of phrase), and in late January, 2000, Kovacs ran the patch through Author. The software responded with an error message.
Kovacs reached out to Escobar. After a few brief exchanges, he agreed to send her the manuscript, which the linguist displayed an eagerness to attempt translating.
“Agreeing to this proposal,” Kovacs said, “was my first mistake. But I was in too deep.”
For the next few weeks he got radio silence from Mexico. During this time, Kovacs resumed traveling for his book tour.
When Escobar reached out again, contacting Kovacs by telephone at a hotel he was staying at in Berlin, it was to inform him that she had met with a number of fellow linguistic scholars in Mexico City and wanted, if Kovacs would agree, to develop a small task-force to translate the text.
“I was desperate,” Kovacs explained, “to get the book translated into English so I could give it to my publisher. I didn’t even want to repeat what I’d managed to do with the first book anymore, I just wanted to deliver on my deadline and wash my hands of the whole thing.”
In order to begin, Escobar told him, she and her team needed to know the manuscript’s origin. Kovacs was in a bind. He understood that it was paramount to an accurate translation that the translators be let in on his secret, but he wasn’t sure he could even trust the professor, let alone the dozens of other scholars who would be involved in the process. Kovacs decided that he had no choice but to tell them everything, but told Escobar that he could only do so once he had met the team in person.
“We’re eager to get started,” Escobar said in an email dated March 2, 2000. Sighting medical complications, Kovacs cut his book tour short and flew direct from London to Mexico City on March 3.
“I had no idea what I was dealing with,” Escobar told me by telephone when I reached her in Mexico City for comment. “Obviously it was some ancient logographic language, but no one on my team could place it beyond having South American origins. Past that, we really had no idea. We’re making a little more headway now, but it’s too soon to tell if we’ll actually be able to decode anything substantial.”
Kovacs’s arrival in Mexico City was kept secret to the extent that, according to him, he slept on a cot in Escobar’s faculty office at the University in order to avoid registering his name at any hotel. Kovacs met with the team of translators on March 4 and, in a discreet conversation with Escobar, revealed his predicament. Escobar agreed to keep the secret of the Author software, but asked Kovacs if he understood that she would have to share the information with her team. He said he did, and the two drew up nondisclosure agreements for everyone on the team to sign. Kovacs canceled the rest of his book tour and stayed on in Mexico City at Escobar’s request to assist with the translation process in a consultation capacity.
“There really wasn’t anything for me to do,” Kovacs said. “Every night I slept on this cramped cot in that woman’s office and every day I would go into the research facility where the team was working and just hang around. No one ever asked me anything. They were all so busy and excited and academic and utterly stupid. I saw clearly, much more clearly than they did, how the whole thing was a fool’s errand.”
In August of 2000, Kovacs was hospitalized in Mexico City with severe ulcers. The following month, he flew back to Moscow to recuperate at home.
“At this point I knew it was hopeless,” Kovacs said. “I knew that Author couldn’t perform the translation, and Escobar and her team weren’t making any headway, and the whole thing was going to kill me.”
Kovacs called Escobar and told her to hang up the project, to which the linguist adamantly refused.
“It had become, for me, a scientific matter,” Escobar explained when I asked her about her transgression against Kovacs’s wishes. “There was no way I was going to stop just because of that man’s defeatist attitude. We had separate interests, and separate problems.”
That same week, Kovacs was awarded three major literary awards. He was absent from all three award ceremonies. This absence, and rumors circulating of his declining health, attracted the attention of the press. Kovacs, unsure of how to proceed and beginning to panic, did not leave his apartment for three weeks, ceasing all communication with the outside world, except his doctor, who continued to fill prescriptions for increasingly more powerful antidepressants and painkillers.
Near the end of September, an anonymous source on Escobar’s team leaked the secret of Opus’s production to the panel of the Berlin Award, and the linguistic mystery of the second manuscript to the Associated Press.
“To keep this secret means to go against everything I believe in,” the whistleblower said in a statement, citing “good scholarship” and “authorial integrity”.
“The funny thing,” Kovacs said, “is that I can’t blame them. In all likelihood, I would have done the exact same thing if I were in their position. I know how difficult it can be to watch one of your peers enjoy success.”
But once the truth about the Author software had been revealed, Kovacs’s success was not long for this world: Every literary prize he had been awarded was revoked, and a debate sprang up as to whether or not the prizes should be re-awarded to the Author software itself (a debate that is still underway at the time of this publication); Kovacs’s publishing contracts were voided, with both publishers citing his knowing infringement of the plagiarism clauses therein, and he was sued for the generous advances they had given him at signing. Kovacs’s busy schedule over the last year had prevented him from spending much of the money, which he readily returned. He weathered all of these setbacks surprisingly well, saying in a public statement that everything he had been forced to relinquish were things he had never expected to attain in the first place.
“All the while,” he told me “I was secretly hoping that the computer science community would sort of save me, that at least they would still recognize my innovation and redeem me of the whole scandal.”
But when Kovacs was fired from the University of Moscow in late 2000, he realized the naivety of this expectation. In January of 2001, his Russian visa was revoked.
Following these losses and his excommunication from the literary and academic industries, Kovacs returned to California. He drifted between family and friends’ homes for a few months, before settling in the San Francisco Bay Area, where the status of his housing and employment remain unclear. When I asked him about his prospects, Kovacs said, “It’s obvious no one will ever hire me again, at least not in any legal enterprise.”
Whether or not this means that Kovacs could – or has – found employment carrying out illicit activities is unsubstantiated, but the demand for competent computer programmers in international espionage has only grown in recent years and it is easy to imagine ways in which a sequencing software such as Author could be extremely useful for predicting the outcomes of something decidedly non-literary, such as military action. One can only imagine that the price a government or foreign interest would be willing to pay for such a software would well exceed even the most robust publishing contract. But this is merely speculation.
A Russian Intelligence raid of Kovacs’s vacated Moscow apartment turned up five laptop computers and one mainframe hard drive, none of which contained any trace of the Author software. Kovacs claimed to have destroyed the program before leaving the country, though it is possible that traces remain.
“When you work on something for so long,” Kovacs said, “it becomes a part of you.”
Even if the Author software no longer exists, the outstanding question now is how the product of its brief existence has already irreparably impacted literature. Both the North American and European publishers of Opus have made public statements claiming that they will no longer print the book, while several libraries around the world have issued contradictory statements that they will continue to make the novel available as an historical and scholarly document. Opus is being used as a sort of Rosetta Stone in the effort to translate Author’s second manuscript, which has become an international endeavor headed by Escobar.
It is, of course, too soon to tell how Opus has actually impacted novels and novelists themselves. Since no human writer is capable of working with the speed and diligence of the Author software, we have yet to see what a novel written post-Opus actually looks like. Surely, some writers will attempt to outcompete the program, while others will see the status of Opus as a defeat and hang up their literary ambitions. Still others will not have read the book at all – some will have managed to miss it altogether – and go on writing with no sense of comparison whatsoever. As Kovacs himself pointed out, it is impossible for a writer to read everything. But I am sure that the works of literature to come – those which result from the reading of Opus and those that do not – will be the ultimate proof of Kovacs’s hypothesis that greatness can be reduced to the most recent entry in a sequence. Or perhaps his theory will be disproved by a deviation from the sequence so great and unexpected it could not be anticipated by the Author software or any other computer program: The result of chance.
Julian Berlinsky, San Francisco
Max Blue writes about the visual arts and modern culture. His criticism has appeared in SF Weekly, the San Francisco Examiner, and Art Practical, among others. His fiction has appeared in Your Impossible Voice and Mount Hope Magazine, among others. He lives in San Francisco.