INSEAD marketing professor Philip Parker has, by his estimation, authored over 1 million books.
His name is on their covers, but he hasn’t actually written them all -- that chore falls to a machine.
Parker has developed a small arsenal of algorithms capable of automatically generating textbooks, crossword puzzles, poems and books on topics ranging from bookbinding to cataracts.
The software isn’t intended to replace Shakespeare or Updike (though a fiction-writing algorithm is in development). Rather, Parker hopes to generate written materials on niche topics and in rare languages that would be economically unfeasible for traditional publishers to produce. His titles on Amazon.com, more than 100,000 in all, range from economic reports ("The 2007-2012 World Outlook for Luggage and Utility Racks," priced at $795) to Unami-to-English crossword puzzles to medical guides ("The Official Patient's Sourcebook on Vocal Cord Paralysis.")
Parker started by generating market reports sold to banks, consulting firms and government trade agencies interested in the sales outlook for, say, rubber or corrugated cardboard. Now, he hopes to use the algorithms to help with language learning and education in developing countries. Thanks to Parker’s automated radio broadcasts, people in parts of Malawi are hearing weather forecasts in the local language for the first time -- and are already changing their farming patterns as a result.
For our “Life As” series, we spoke with Parker about authoring with an algorithm, why dissertations could be automated and how watching television could change in drastic ways.
Will I be obsolete soon? What are the implications of your software for writers? There are very few implications for writers because we’re covering areas that writers don’t or can’t cover. When you’re looking at small languages, the population of speakers is so small that there might not be people with the expertise in science or agronomy to write optimal planting strategies for maize in the local language. Every title we create is educational. All of these projects seem very diverse, but really they’re all in same vein: we’re using automation to reach areas that wouldn’t have been served otherwise. And it gives people the opportunity to see or experience content they wouldn’t have been able to otherwise.
But surely we can outsource some existing tasks to your algorithmic authors? Already there are algorithms writing articles for Forbes, for example. There are many forms of writing that are common, but also very formulaic, such as annual reports or economic studies. In those areas, people would probably be relieved not to have to write those kinds of things because they are mundane and drudgery.
If you want to wax philosophical, you could see these algorithms as an enabling tool. There are a lot of people who want to be writers who stumble at a blank page. You could imagine an algorithm that could give writers a first draft or a starter kit, so it could enable people to be more prolific in their writing.
How might your technology change how -- or what -- we write? One of the areas I’m working on is, can we create a doctoral-level thesis that’s fully automated -- to save the pain of four years of a Ph.D. program -- and still come up with an original conclusion? If we could do it in an automated way, we’d increase the speed of discovery. As time goes on, someone might be able to say, “I’d really like to do a doctoral thesis on X and Y with a Z perspective,” and with a few clicks, it will have be authored.
What changes do you foresee in how we consume content? People will be seeing content they wouldn’t otherwise have seen in their language. But this technology is nowhere near maturity – it’s like the automobile industry in the late 1800s right now. One hundred years from now, things I can’t even imagine will be available. When you watch television, for example, there will be a channel specifically for you generated on the fly using your particular profile. It’ll be rendered in real time before you even watch it, so you may be the first person to ever see that program, but it will be what you most want to see.
We’re far from that. But is it technically feasible? Probably.
How does your software actually write the books, poems and puzzles you produce? There’s no single piece of software, per se. Each genre we do has its own very customized algorithm that’s been created. For international trade reports, we use econometrics. To write poetry, we use graph theory -- we have a semantic web of knowledge, which took us three or four years to create, and we use a trained database to mimic the human mind in terms of word usage.
What can human authors do that these algorithmic authors can’t? The toughest thing will be the creation of genres themselves -- A computer would have a hard time inventing cubism, for example. Formula creation is magnitudes harder for computer algorithms than actually executing within a formula. But once the genre is created and the formula is known, then the computer can do the repetitive task of executing within the genre.
This interview has been edited and condensed for clarity.