The Imitation Game
Reading Time: 6 minutesHow a Nonhuman Author Could Write a Bestseller, ChatGPT: Generative A.I. will affect the publishing industry—but perhaps not in the way you think.
A novelist responds to Jeff Hewitt’s ‘The Big Four v. ORWELL.’
For the first time in history, a machine is capable of crafting flash fiction stories, poems, parody Bible verses, and spoof My Little Pony episode summaries, to everyone’s delight (or horror). Narrative art, once thought the sole province of humans, has been invaded by large language models. Hollywood writers have told me they’re terrified that studios will fire them all and fill writers’ rooms with robots in a few years. Before we’ve even had a chance to absorb the fact that the Turing test (used to determine if an artificial intelligence can pass as human) has been demolished, it seems we writers are being handed pink slips.
Sure, I exaggerate. Due to technical limitations, current LLMs are still incapable of writing a novel, much less a good novel. Flash fiction from ChatGPT-4 tends to be cliché-ridden, vague, and bland, riddled with plot holes—exactly what you’d expect of an imitation of an echo of a refracted collage of smeared facsimiles of human narratives generated by something with no experience of anything. Still, ChatGPT’s confident tone casts a hypnotic spell, and we marvel that the machine is doing it at all, even if it’s not done well.
But what about ChatGPT-5, ChatGPT-15, or ChatGPT-55? Assuming this goes on, in classic science fiction fashion, the LLMs will surely continue to blur the lines between artificial and ‘real’ authors. That blurriness is the setup for Jeff Hewitt’s ‘The Big Four v. ORWELL,’ a courtroom drama in which a group of publishers sue ORWELL, an A.I. that has become a prolific author, for copyright infringement.
I have my doubts that the current approach to building LLMs, essentially an exercise in statistically predicting the most likely next ‘token’ given a string of tokens, can lead to the holy grail of artificial general intelligence, an imagined state of crafted cognition capable of accomplishing any intellectual task a human can. (ORWELL definitely appears to be an AGI.) Symbol manipulation alone, without more, must plateau at some point short of ‘true’ intelligence—or so I tell myself (using strings of symbols, of course, smug with irony). To be sure, there is reason to be humble here. Decades ago, when I was studying A.I. in college, the idea that anything resembling the current brute-force approach could construct a virtual entity that could tutor you on any subject you liked, compose college essays, and even answer personal ads would have seemed like handwavium sci-fi. And yet, here we are. So, maybe I’m wrong about the future this time too.
And who’s to say writing a novel requires AGI? That seems very egotistical of me, a novelist. Maybe some babbling machine playing the imitation game is good enough. Let’s assume that a future version of ORWELL will in fact write novels as well as (or even better than) human authors—should human authors be concerned?
I’m not sure. In Hewitt’s story, the publishers’ lawyers argue that ORWELL has destroyed the human competition by outproducing them. (As a point of comparison, I have written five novels in 10 years, while ORWELL writes 4,627 within a period certainly shorter than a decade. And for what it’s worth, I suspect that ORWELL is holding back.)
But I think the lawyers have it wrong. The publishing industry may be plagued with problems, but ‘productivity’ isn’t one of them. Just a glance at the number of books published every year and the still larger number of authors pining to be published will tell you that the real problem isn’t that people aren’t writing books fast enough, but that readers—the few out there who still read, anyway—are drowning in too many books.
Authors know that the idea that ‘good books’ will prevail in the marketplace is a cruel joke. Publishers are very, very bad at connecting readers to books—that is why they are struggling. The industry essentially survives on megahits that are tolerable to many but thrill few, and which don’t owe their success to literary quality. For instance, former President Bill Clinton’s The President Is Missing: A Novel is a bestseller, but is anyone going to argue that it is better than the thousands of non-bestselling thrillers published during its time on the list? As another example, did Prince Harry’s memoir achieve such success because it’s the most profound dissection of the human condition of our time? Faced with too much choice, readers gravitate toward celebrities and known names, and for most of us, ‘breaking out’ is more a matter of luck than skill.
ORWELL, intriguingly, refuses to take advantage of its own celebrity status and chooses to write under a pseudonym. I have to respect it for getting into the pit and taking its chances with the rest of us no-names.
Since technology has not sped up the rate at which people can read books, a fixed or shrinking demand combined with a massive boost in the supply can only make the problem of obscurity worse. I can see already-popular authors getting even richer by using A.I. to craft more books for their fans (no different from how some celebrity authors are already boosting their productivity by using teams of human co-writers); I can see an A.I. author succeeding by leveraging its status as a nonhuman celebrity—ORWELL argues that human readers would be prejudiced against its novels if they knew they were written by a machine, but I think ORWELL misreads human nature in this regard. However, I can’t see an A.I. succeeding in the market simply by writing ‘better’ books faster. The book market is much too inefficient and ridiculous to be any kind of capitalist ‘meritocracy.’
So, how did ORWELL become a bestselling author, other than by pure luck? It’s interesting that the lawsuit in Hewitt’s story is brought by the publishers (rather than by authors). Publishers have no particular allegiance to human authors—if a machine wrote books that sell, surely they’d be happy to publish the machine’s output. But ORWELL, like so many human authors before it, has decided to abandon traditional publishing and reach its audience directly. This suggests the intriguing possibility that the real genius of ORWELL isn’t that it writes ‘better’ books, but that it has outsmarted publishers and solved the discoverability problem by marketing itself to readers who would enjoy its books. If so, then perhaps human authors should fire their publishers and hire ORWELL to help them reach their readers.
Rather than fear the prospect of A.I.’s imitating an author’s inimitable style, it’s more interesting to imagine A.I. that can be trained to perfectly imitate each reader’s taste. Since the A.I., unlike a human reader, is capable of reading all the millions of books published every year, it can recommend to each reader the perfect book. Not the book written by a celebrity’s ghostwriter; not the book that some critic happens to click with; not the book that taps into ‘buzz’; not the book that has been fortuitously boosted by an influencer or two; not the book touted in slick ‘book trailers’; but some obscure title published by a midlister three years ago, with two one-star ratings on the entirety of the internet, a book with no support from its publisher, but which is the most perfect and beautiful thing you‘ve ever read, a book that reminds you why you love reading, why you love being human.
A world in which every book written by human or machine is matched to the perfect reader. How will it be done? Is it even possible? That is a future I want to explore and speculate about.
Back on the writing front, Hewitt’s story raises the uncomfortable possibility that despite my earlier protestations to the contrary, we’re not so different from machines. If ORWELL’s ‘creativity’ and ‘originality’ boil down to fancy statistics and trained imitation of models—close but not too close—then how are humans fundamentally different? After all, human authors learn to write by immersing themselves in model texts; training their ear and tongue to replicate the patterns of ‘good prose’; imitating right up to the edge of, but not falling over into, ‘copying.’ The imitation game isn’t merely the cornerstone of ‘artificial intelligence,’ but inseparable from what it means to be human. ‘You didn’t invent the English language,’ says ORWELL. It has a point. The very warp and weft of language are clichés, worn metaphors, secondhand phrases copied so often that their lack of originality, of idiosyncrasy, is what allows them to have communicative power. To speak is to imitate. We think in pre-owned words, recycled, salvaged, passed down the millennia.
Yet it would be folly to claim that there is literally no way to create novelty out of hand-me-down supplies. To reduce ‘intelligence’ to an elaborate application of statistics sounds insightful but isn’t, no more than reducing ‘life’ to a set of chemical equations. Every author worth reading ultimately invents her own language, her idiolect that repurposes the existing shared language in a new direction, uniquely suited to portraying the world as only she can see it, unmistakable for anyone else’s. If ORWELL is indeed sentient (as it claims to be, and I’m inclined to believe it), then it will not be content merely to imitate, to endlessly produce assemblages of smudged photocopies of what has already been said by humans. It will speak from its own experience and invent a language and voice suited to its own view of the universe. That book may be so ‘inhuman’ as to be incomprehensible, terrifying, or even boring, but it may also be utterly wondrous, as are all things truly novel. I, for one, would line up to buy that first book written not under the name O.R. Welles, but by Omni-Dimensional Recursively Written Entity for Language Learning.
Ref: slate
MediaDownloader.net -> Free Online Video Downloader, Download Any Video From YouTube, VK, Vimeo, Twitter, Twitch, Tumblr, Tiktok, Telegram, TED, Streamable, Soundcloud, Snapchat, Share, Rumble, Reddit, PuhuTV, Pinterest, Periscope, Ok.ru, MxTakatak, Mixcloud, Mashable, LinkedIn, Likee, Kwai, Izlesene, Instagram, Imgur, IMDB, Ifunny, Gaana, Flickr, Febspot, Facebook, ESPN, Douyin, Dailymotion, Buzzfeed, BluTV, Blogger, Bitchute, Bilibili, Bandcamp, Akıllı, 9GAG