The first full-length Korean novel written by artificial intelligence (AI) hit bookshelves in Seoul last month. The World From Now On, published by Parambook, was written by an AI author named Birampung under the direction of novelist and mathematician Kim Tae-yon.
The 560-page book tells the story of five people — a disabled amateur mathematician, a math professor and entrepreneur, a psychiatrist, an astrophysicist and a Buddhist monk — who are drawn to each other in their individual quests to understand the meaning of human existence.
Kim has previously published five full-length novels and two mathematical novels. In publishing his latest work, Kim said he picked the storyline, background and characters and delegated most of the actual writing to Birampung, a name that refers to a fierce storm that strikes at the beginning and end of the universe’s creation. Kim was reluctant to share the details of the technology involved, claiming it was sensitive information. But he compared his job to that of a movie director, saying he gave “orders to Birampung” and after checking the results made adjustments as necessary for a “retake.” The AI writer was fed with around 1,000 books, including some of Kim’s own, more for it to learn his writing style. Birampung’s writing virtually needed no grammatical or language correction, or so says Kim. The output was very clear and used certain well defined techniques that had been coded into the writing guide that the AI used. Most of all, the AI displayed an incredible ability to draw specific details from its acquired knowledge from the thousand books that the engine had been made to “read”.
To be fair though, Birampung too experienced writer’s-block enroute, and would sometimes go absolutely quiet for days. Also, the AI would often go into too much detail or use expressions that went overboard or were slang or sometimes completely outdated. That is when his minder, Kim, had to step in and say, “Hey, that’s not what I meant!” and get corrections effected mid-course. The book eventually took seven years to complete.
Parambook, the publisher, claims that the book is the first “real” full-length novel written by AI after more crude attempts in the past in Russia in 2008, and in Japan in 2016. And now a Japanese translation of Kim’s book is already in the pipeline. Hopefully, other language translations will follow.
So, is writing an imperilled profession? Will technology displace the very art of the wordsmith and will the Shakespeares and Byrons of tomorrow all live in coded engines? News from another related domain is equally alarming.
At the annual American Crossword Puzzle Tournament (ACPT), an AI system built by a team of UC Berkeley researchers emerged the winner, beating all human competitors by quite a mile. This year’s tournament was held virtually and had a record number of 1,287 contestants. While the tournament was open only to human competitors, the AI system, named Dr Fill, that “won” the tourney, competed unofficially.
Dr Fill solved the final puzzle in 49 seconds, more than two minutes faster than the top human contestant — which left other competitors both amazed, and somewhat dismayed.
Another profession that seems under threat is from my own world: Copywriting. Those who do crosswords regularly would know that there are two steps to solving a crossword. First, you figure out possible answers for each clue. Then you take all possible answers and figure out which ones go together on the grid. The AI system, Dr Fill, seems to have done both steps with ease and accuracy at the ACPT tournament. So the basic ingredients of good copywriting — good vocabulary and the right deployment of the words — seems to have been figured out by the new AI discovery. Who knows how long it will take to learn to describe things, write sentences and curate copy to a client’s brief! Kim’s Birampung already seems almost 90 per cent close.
The inevitable question is how soon can you and I get to use AI to write? Well, perhaps we can do it pretty soon, using an entity called GPT-3. It’s an artificial neural network with over 175 billion parameters — think of it like an artificial brain with the computing power of 175 billion connections (if it makes you feel any better, you probably have around 125 trillion synapses in your own brain, for comparison). GPT-3 is a natural language processor, which means it’s trained to try to complete any prompt that it’s given. Its training data is basically the entire internet, so given a prompt, like a few paragraphs of text, it guesses what comes next. These guesses get GPT-3 to write in all sorts of styles, oftentimes as convincingly as a real human author.
Am I elated or disappointed? Honestly, I’m just a little creeped out.
The writer is managing director of Rediffusion