Indy and the Library of Babel

When Alfred Lang died two years ago, obituaries mostly covered his stock market achievements. For in a lifetime Lang had made over $120 billion in profits starting in late '90s and through the Dotcom bust of '00s and the recession of '08. This success got covered in books, magazine articles (many in this publication) and two successful films. And then as if to balance it out, the obituaries allotted a requisite amount of print space to Lang’s 'minor' eccentricities (by billionaire standards): owning a 150 foot yacht and making angel investments in an entity housing data servers in the Artic (to save on cooling costs). But none mentioned the passion that consumed Lang for most of his life: his search for lost books. And absolutely no one knew about the AI project he funded in the last six months, called Indy. 

Yes, after Indiana Jones. 

Now Lang’s family is suing Thubron Software, the company that built Indy at the behest of Alfred Lang. In its claim the family says Thubron Software deliberately misled an old man into thinking that the idea was plausible, selectively presenting data in a way that hid the fact that success was near zero. The family could win a settlement of nearly $450 million if it wins, a sum that might send Thubron Software into bankruptcy. In the last few weeks I met and followed the main protagonists of this tale to understand an old man's bizarre search for lost books. 

My journey began in the city of Gilroy, south of San Jose, California where I travelled to meet Alfred Thubron, chairman and CEO of Thubron Software. Thubron is the son of an Englishman and an Indian immigrant. His parents moved to USA from Britain and India in the late 1990s and met while studying at Berkley around the time Alfred Lang’s success had begun to feature him on magazine covers. 

“It's hard to believe now, but I hadn’t heard of Alfred Lang until he approached me with the proposition for using AI to find lost manuscripts,” said Mr. Thubron. We were settled in his manicured garden with a table full of Indian food laid out before us. Mr. Thubron is a heavy set man in wire rimmed glasses who prefers dressing in khakhi shorts and round neck T-shirts.  

“ After we introduced Agile, a lot of investors were interested in investing with us,” Mr Thubron said. Agile is a Large Language Model developed by Thubron Software specifically for writers. It is designed to analyse any writing and tell if the style is closer to someone else's. "Suppose you are a writer trying to write a story about fly fishing, Agile will not just tell you that your idea is like Hemingway's, it will also tell you which of Heminway's story plot and style is closest to yours. The idea is that once you know who you tonally resemble you can better understand their style and build over this."

The idea also had other uses, like catching plagiarism with greater efficiency and for this Mr. Thubron was often approached by various universities and businesses.

“After a conference in Las Vegas Mr.Lang asked us to demonstrate Agile for him.” At first Thubron didn’t want to do it. He didn't see how Agile could help a stock investor. "And we were getting good reviews in the press and already had offers from other investors to put money in his company. So I wasn't looking for an investment." Yet he chose to meet Lang because Lang told him that “he had a personal collection of over 5000 books now out of print and needed help finding more.”

“I was intrigued. I have always been a book lover and when I heard that there was this man, who specifically collected lost books, I wanted to know more. I figured even if we don’t work together, I might get a new friend out of this.”

During the meeting it became clear to Thubron that Lang was obsessed with the search for lost books. “ I realised that I was the one being presented material here,” Mr. Thubron recalls, “Lang showed me title after title of great lost books he had helped to find. Books no one had heard of. He showed me first editions of books that had been popular in 1909 and were now forgotten.”

I looked up the list of books popular in 1909. Lang was right. The best seller in 1909 was a book called The Inner Shrine written anonymously by a clergyman called Basil King. The book is now entirely forgotten. A first edition copy was on sale in the year 2012 and was picked up by Lang. 

In that meeting Alfred Lang also told Thubron about the great library of Alexandria. Thubron like most people had a vague idea about the destruction of the library, but it was clear Lang was into the weeds. “He showed me a probable list of 50 plays by Euripides known to have been lost in the fire. And the Dialogues of Aristotle written by his student Theophrastus. I did not know any of this at the time.”

“What did you think of Alfred Lang then?” I asked Thubron that evening when we were walking his dogs. He had to stop and think about this. 

“You know I have never asked myself this question,” he said. “Physically, I was struck by how short Lang was (laughs) but beyond that I thought of him as being a little eccentric, like I expected a billionaire to be.”

If there is an offence Lang would have taken, it would be this description of him as an eccentric. Most of his life Lang strived to maintain an almost middle class normalcy. Being born in Vietnam after the war, he had witnessed extreme poverty very early in life and this made him an enlightened capitalist. This story has been documented elsewhere (links below), but not his obsession with lost books that also came from this time.

I asked his son, Julian about it. 

Julian Lang lives in a Crazy Rich Asian style mansion in Singapore. He stays there with his wife and two kids and runs the family Arts Centre set up in honour of his mother. He is the exact opposite of Alfred Thubron in his choice of clothing and was dressed crisply in fine white shirts and dark trousers. I met him in his oak panelled office on the second floor of the Wendy Lang Arts Centre. 

“My father's fascination with lost books started after a fire destroyed the library at his school in Vietnam. Being one of the few kids who loved reading, he felt the loss deeply. This was a time when replacing books was not a simple matter of ordering them online or downloading ebooks. If you had a fire in the library, you lost everything. Probably permanently.”

In an interview to the New York Times book review podcast, Lang talked about the aftermath of this loss. I listened to a recording of this interview on my way to Singapore and played it to Julian. 

“It was as if I had lost the reason for my existence,” Lang says in a high pitched soft voice, “You see books were my only source of mental nourishment. They told me how the world worked, how vast and beautiful it was and how full of adventure it was. After the library fire it felt like someone had left me holding the embers of the world.”

“How did you cope with it,” asks the interviewer. 

Lang went on.

“My grandfather told me that books are never lost. They travel to the ends of the world and are archived in the Library of Babel”

Julian smiled. I asked him if this was the first time his dad had mentioned The Library of Babel in public. 

“I think so,” he said. “But we knew it from childhood.”

It is important to introduce the Library of Babel here. The Library of Babel is a short story written by the Argentinian writer Jorge Luis Borges, first published in 1941 as a part of his short story collection The Garden of Forking Paths. In the story the library (otherwise known as the universe) contains something called Hexes, a series of infinite hexagonal rooms with five bookshelves on each wall , each shelf with 32 books and each book with four hundred and ten pages each. 

The Library of Babel is not a real library. It is an idea about books. Books here are randomly generated, from a combination of 25 symbols - 22 letters of alphabet, coma, period and space. Since most of language is the combination of letters and symbols, this endless random stringing together of letters and three symbols could theoretically produce every book ever written, being written and that will be written in the world. 

I asked Julian what he thought about this. 

“It is a good story, but not a workable idea. Theoretically even if the random combination of 25 letters can produce your biography and my biography, there will be infinite versions of this making it almost impossible to find anything concrete.”

And yet, his father was fascinated with the idea of Library of Babel. For years he discussed this with anyone who listened.

“His eyes lit up when he spoke about it,” says Julian, “he brought it up with friends when they came over. The hook was always that somewhere in the library was a book on the listener’s life, already written.”

But it was not predicting future that interested Lang. The real purpose of the Library of Babel for him was that somewhere in the library were copies of books lost to humanity. Like Cardenio, William Shakespeare’s lost play based on a character in Don Quixote. 

“I think for him Cardenio was the gold standard of lost books. A book at the confluence of both Shakespeare and Cervantes, he thought it was one of the greatest lost books of all time,” Alfred Thubron told me. “And when we met the next time, he wanted me to build him a version of Agile that could help find Cardenio in the Library of Babel. This meant that we would be building a software that can find any lost book in the Library of Babel.”

Ever since the beginning of the World Wide Web there existed websites that claimed to have all the books in the Library of Babel. At a simple level, all it took was to write a code that could combine the 25 characters randomly. Alfred Lang got a version of the Library installed at his home very early in the Information Age. But almost all the books produced by the program were gibberish, a random series of letters strung together that made no sense. Lang spent a lot of money tweaking these programs and got nowhere. Take Cardenio. We know Shakespeare based the play on an episode of Don Quixote, we know it was probably written sometime between 1611 and 1613, we know it was set in Sierra Lorena area of Spain and because it had been adapted by a playwright called Lewis Theobald as Double Falsehood in early 1700s we know the plot. Each one of these knowns is a variable that should make the search for Cardenio easier. It was simply a question of adding greater number of certain variables to narrow down the search.

I asked Thubron why then this was still hard. 

“Even with these constraints, there exist in the library a dizzying number of books that meet all the criteria and are still not Cardenio. The search with these constraints yields millions of books, one of which could be Cardenio. The only way of knowing which one is by physically looking at each one.”

For fifteen years Alfred Lang tried every new approach and every new tool that came into the market to search the Library and failed. I asked Julian how much his dad had spent on this elusive search. 

“I don’t know,” he said. “Dad’s Library of Babel projects were strictly done under his supervision. And till he lived, no one was allowed to question his financial decisions.”

“If you had to estimate..” I insisted. 

“Perhaps hundreds of millions,” he said with a sheepish smile. “But a lot of this was to Thubron Software.”

After that meeting in Vegas, Alfred Lang approached Thubron with a proposition. 

“He wanted us to create on overlay on Agile. You see the world had seen how quick Agile was in understanding the style of any writer and put a score on any other’s work to tell us how close it was to the writer. So could it not scour through millions of Cardenio possibilities and help narrow down the search for the true copy?”

What is an overlay? I asked. Thubron explained. “It is a specialised version of Agile modified to do one task. In this case to run through the hundreds of possible versions of a lost book and narrow down the search to a few. We called this overlay Indy - after Indiana Jones - searching for lost treasures - in this case - books.”

What happened next is under dispute. I heard Mr. Thubron first. “When we first ran Indy it eliminated nearly 70% of the false positive Cardenios on the basis of simple errors of fact, spelling or style. When we reported back our results to Alfred Lang he was convinced that with Indy we were closer than ever to discovering a lost play of Shakespeare. We have emails from him about this. He was ecstatic with our progress.” 

Of course finding Cardenio would have been a magnificent coup for Thubron software. It probably would have been a watershed moment for all AI. Within a year of its existence, artificial intelligence would do what no human has been able to do in over 400 years. 

This is the case made by Thubron software who say Alfred Lang willingly paid tens of millions of dollars in server time knowing Indy was closer than ever in isolating the section of the Library of Babel that could house the original Cardenio. 

“Of course once the case was proved with Cardenio, we could replicate the results for any writer. Alfred Lang knew this. He even lined up the projects we could take up afterwards with Indy.”

Julian disagrees with this proposition. “They were never close to finding Cardenio. Even after Indy eliminated 70% of the possibilities what was left was a dizzyingly large number of books that could be Cardenio. If I remember correctly the number is still in millions.”

The correct number is ten billion. I asked Thubron if this was right. “Yes, it is.”

He insists that the number appears large in isolation. “We started with a number many times that and had eliminated 90% of alternatives. With Indy learning to spot greater and greater patterns that eliminated more versions, we estimated the search to find the true Cardenio was a matter of months, if not weeks.”

And this the point when Alfred Lang died and the obituaries mentioned at the start of the article appeared in newspapers. Soon the family stopped paying the bills for Indy’s costs. 

“And in August last year we got sued,” Thubron told me. 

“How close do you think you got to finding the lost Shakespeare?” I asked Thubron before leaving Gilroy. 

“Only a few weeks,” he said. 

I asked him why he didn’t fund the rest of the project himself, given they were so close to a result. 

“Because Alfred Lang had stipulated that all results, code and material related to Indy belong to him,” he told me, “This means Julian and his family call the shots on this.”

I asked Julian if it wouldn’t be a tribute to his father to find Cardenio. 

“Alfred Thubron is self delusional if he thinks that he can find a lost Shakespeare in matter of weeks. We need to step back a little here and see what we are talking about. The idea that a random combination of letters can become Shakespeare! Isn’t that a disservice to all art when a purely mechanical process can create Shakespeare. I find this line of thought repulsive.”

In the book review podcast Alfred Lang says that as a ten year old he asked his grandpa how to find the Library of Babel.

“We cannot get there my son,” said the old man, “The Library is a dark place where all light runs out things becoming endlessly heavy. Unless, you become one with the idea.”

I asked Julian if he knew what he meant by becoming one with the idea. 

Julian Lang got up, shrugged his shoulders and said, “I think it means you become one with the idea of permanent loss. Some things are lost forever.”