Article de Noam Cohen, paru dans le New York Times du 14 avril (et reproduit dans le supplément hebdomadaire du Monde de ce week-end). Où il est question d'une nouvelle méthode de fabrication des livres, assez édifiante ou effrayante, c'est selon. Ecrire un livre ? Un jeu d'enfant, vous copiez-collez à partir de vos recherches internet, vous compilez à la va-vite, vous enrobez le tout d'un sommaire et d'un titre accrocheur, et le tour est joué... C'est cette méthode bête comme chou qu'a adoptée un certain Philip M. Parker, professeur de management dans le civil et qui systématise ce tour de passe-passe (battant au passage le record de référencement d'ouvrages publiés), plutôt dans le domaine scientifique. Poètes et zélateurs de la paresse et de la flânerie s'abstenir...It’s not easy to write a book. First you have to pick a title. And then there is the table of contents. If you want the book to be categorized, either by a bookseller or a library, it has to be assigned a unique numerical code, like an ISBN, for International Standard Book Number. There have to be proper margins. Finally, there’s the back cover.
Philip Parker is now turning his efforts to video.
Oh, and there is all that stuff in the middle, too. The writing.
Philip M. Parker seems to have licked that problem. Mr. Parker has generated more than 200,000 books, as an advanced search on Amazon.com under his publishing company shows, making him, in his own words, “the most published author in the history of the planet.” And he makes money doing it.
Among the books published under his name are “The Official Patient’s Sourcebook on Acne Rosacea” ($24.95 and 168 pages long); “Stickler Syndrome: A Bibliography and Dictionary for Physicians, Patients and Genome Researchers” ($28.95 for 126 pages); and “The 2007-2012 Outlook for Tufted Washable Scatter Rugs, Bathmats and Sets That Measure 6-Feet by 9-Feet or Smaller in India” ($495 for 144 pages).
But these are not conventional books, and it is perhaps more accurate to call Mr. Parker a compiler than an author. Mr. Parker, who is also the chaired professor of management science at Insead (a business school with campuses in Fontainebleau, France, and Singapore), has developed computer algorithms that collect publicly available information on a subject — broad or obscure — and, aided by his 60 to 70 computers and six or seven programmers, he turns the results into books in a range of genres, many of them in the range of 150 pages and printed only when a customer buys one.
If this sounds like cheating to the layman’s ear, it does not to Mr. Parker, who holds some provocative — and apparently profitable — ideas on what constitutes a book. While the most popular of his books may sell hundreds of copies, he said, many have sales in the dozens, often to medical libraries collecting nearly everything he produces. He has extended his technique to crossword puzzles, rudimentary poetry and even to scripts for animated game shows.
And he is laying the groundwork for romance novels generated by new algorithms. “I’ve already set it up,” he said. “There are only so many body parts.”
Perusing a work like the outlook for bathmat sales in India, a reader would be hard pressed to find an actual sentence that was “written” by the computer. If you were to open a book, you would find a title page, a detailed table of contents, and many, many pages of graphics with introductory boilerplate that is adjusted for the content and genre.
While nothing announces that Mr. Parker’s books are computer generated, one reader, David Pascoe, seemed close to figuring it out himself, based on his comments to Amazon in 2004. Reviewing a guide to rosacea, a skin disorder, Mr. Pascoe, who is from Perth, Australia, complained: “The book is more of a template for ‘generic health researching’ than anything specific to rosacea. The information is of such a generic level that a sourcebook on the next medical topic is just a search and replace away.”
When told via e-mail that his suspicion was correct, Mr. Pascoe wrote back, “I guess it makes sense now as to why the book was so awful and frustrating.”Mr. Parker was willing to concede much of what Mr. Pascoe argued. “If you are good at the Internet, this book is useless,” he said, adding that Mr. Pascoe simply should not have bought it. But, Mr. Parker said, there are people who aren’t Internet savvy who have found these guides useful.
It is the idea of automating difficult or boring work that led Mr. Parker to become involved. Comparing himself to a distant disciple of Henry Ford, he said he was “deconstructing the process of getting books into people’s hands; every single step we could think of, we automated.”
He added: “My goal isn’t to have the computer write sentences, but to do the repetitive tasks that are too costly to do otherwise.”
In an interview from his home in San Diego and his offices nearby, Mr. Parker described his motivation as providing content that the marketplace has otherwise neglected for lack of an audience. That can mean a relatively obscure language is involved, or a relatively obscure disease or a relatively obscure product.
Take, for example, the study of bathmats in India. “Only one person in the world may be interested in that,” he conceded, “probably a strategic planner for a multinational that makes those.” But he points out that once he has trained the computer to take data about past sales and make complex calculations to project future sales, each new book costs him about 12 cents in electricity. Since these books are print-on-demand or delivered electronically, he is ahead after the first sale, he said.
His company, the Icon Group International, is the long tail of the bell curve come to life — generating significant total sales by adding up tens of thousands of what might be called worst sellers. For example, a search at the Galter Health Sciences Library of the Feinberg School of Medicine at Northwestern University found half a dozen Icon books, mainly in the library for patients and their families.
Icon is “a very innovative and interesting example of print on demand,” said Kurt Beidler, a senior manager at Amazon.com who runs the publishers’ services for BookSurge, Amazon’s print-on-demand company. “A lot of examples of print on demand take older books and bring them back — really acting as a supply-chain tool. In this kind of business, it’s a new business, using this capability to introduce new material to customers.”
Mr. Parker compares his methods to those of a traditional publisher, but with the computer simply performing some of the scut work. In an explanatory YouTube video, Mr. Parker shows a book being created. The computer is given an assignment — project the latent demand for antipsychotic drugs around the world, based on the sales figures in the United States.
“Using a little bit of artificial intelligence, a computer program has been created that mimics the thought process of someone who would be responsible for doing such a study,” Mr. Parker says. “But rather than taking many months to do the study. the computer accomplishes this in about 13 minutes.”
An editor picks the years to be covered, but the computer picks the optimum model for extrapolating sales in various countries, and in alphabetical order produces a chart for each country. “It will then open a Word document and export the information into Word just like a real author would out of their minds, so to speak, or spreadsheets,” he says.
Artificial intelligence researchers say computers are far from being what the general public would consider authors.
“There is a continuous spectrum, also known as a slippery slope, between a program that automatically typesets a telephone directory and a program that generates English texts at the level of variety you would expect from a typical human English speaker,” said Chung-chieh Shan, an assistant professor in the computer science department of Rutgers. “The former program is easy to write, the latter program is very difficult; in fact, the holy grail of linguistics. Like Mad-Libs, Parker’s programs probably lie somewhere between the two ends of this spectrum.”
Mr. Parker has lately taken to lighter fare intended to educate. He said he had invested “up to seven figures into the animation business” for word-based video games and animated game shows that will teach English to non-English speakers. YouTube has many examples of these games, which have computer- generated scripts.
A low-tech version of those games are the thousands of crossword puzzle books Mr. Parker has made in about 20 languages. The clues are in a foreign language and the answers are in English. The computer designs the puzzles and ensures that the words become harder as one progresses.
As part of his love of words, and dictionaries in all languages, Mr. Parker said he has taken to having his computers create acrostic poems — where the first letter of a series of words spells a synonym of those words, often to ironic effect.
Of course, one of the difficulties of generating a hundred thousand poems is stepping back and assessing their quality.
“Do you think one of them is Shakespeare?” he was asked.
“No,” he said. “Only because I haven’t done sonnets yet.”
