“Biggest” News in Books: Artificial Intelligence
I’ve now officially read an AI-assisted book on submission, have heard whispers of a partially AI-written novel that will come out from a major publisher, and have been generally swamped with water-cooler conversation about AI and its potential to upend the order of work and life. Despite the increased level of conversation and pace of innovation, my views about the technology remain about the same as they were in March. Primarily, that the ways machines can learn to exploit human behavior using AI is probably far more valuable than the content AI can generate.
Armchair Analysis of Where AI is Going
The best person to read who talks about AI and is immersed in this stuff but writes in a way that is both fun and that we regular people understand is Ryan Broderick of the Garbage Day newsletter. All credit to him as probably the biggest source of my two conclusions to where we are headed:
1) AI will probably be used against humans passively in a set-it-and-let-it-work-its-dark-magic way more so than AI will be used by humans in an active let’s-see-what-new-Grimes-music-we-can-create way. In fact, we’ve seen this with TikTok and across social media already, which has been manipulating our lives for far longer than any of these flashier call-and-response AI tools have been “threatening” to take our jobs. TikTok has much more quickly transformed daily lives than chat-based AI—I’ve now lost count of how many times TikTok videos or creators or trends are referenced in any given casual conversation. It takes a good portion of our day away, rewires our brains, and is legitimately responsible for shifting buying habits (like in books!).
2) The endpoint for AI, and the pretty obvious hope for Microsoft who have integrated the technology into Bing—their search engine that has been functionally dead for years until now—is that it will supplant Google search as the new standard for asking questions and seeking answers on the Internet (Leaked Google documents show they are scared of this possibility; credit again to Garbage Day for this source). A better search engine seems great on the surface: wouldn’t it be nice to get more dynamic search results to a question like “please list five peer reviewed academic papers on the bioethics of Crispr with a summary of their conclusions”? AI’s huge advantage is that it is more generative than informational, which takes a lot of extra work out of scouring through top search results and drawing your own conclusions, but buyer beware: faulty citations and outright lies abound.
But neither of these more likely transformative outcomes of AI are nearly as fun to talk or think about as what grabs headlines. Whether it is the WGA strike or novel submissions, people want to know:
Can AI Create Art?
This is invariably the first question that is asked when a new technology comes along: can AI write Shakespeare or compose Mozart or generate a new Van Gogh painting (these are still our three most cliché and generic examples of artists in their fields, sorry). Which is just another euphemism for can it replicate and replace what makes us human? Whether it is dishwashers, robots, or touch screen kiosks, this question will always be at the heart of fear and our existential conversations about technology.
From the frontlines of a creative industry, it’s pretty clear the answer is somewhere between “no” and “not any time soon.” Technologists will always meet this answer, especially when it comes from people in the arts or journalism, with an eye-roll and Moore’s-law argument about how quickly the technology will improve. But while the singularity may be inevitable, it does often feel like the rapture: everyone selfishly thinks it will be in their lifetime and so you can always find someone to predict it coming soon.
My experience with AI and the creative materials these tools can produce, is that it is very good at achieving a 5th to 8th grade level of competency. AI feels like magic when a chat bot spits out complete and coherent strings of sentences in seconds. And this instantaneousness makes you feel, for a second, like the technology could replace us. The innate truth and logic we follow in this moment is that even if you took a smart person and sat them down to write an essay or story about some random thing, it would probably take them a much longer time to produce something of similar quality. A human does not have all facts at their fingertips and is not trained in a programmatic way to compose in regimented sentences—when you ask someone to write a love letter in Didion’s style, they may want to go read some Didion again before making an attempt. All this makes you say “wow, we cannot compare to machines.”
Upon further interrogation this magical thinking falls apart. It’s important to remember that this was probably the same emotion people experienced when calculators became available for the first time. But while computers are great at math, they’ve been far less successful in creating mathematical theorems. For anyone charged with true creative work, I suspect this analogy to math will hold true when it comes to AI and art. Yes, sometimes the best solution to a writing problem really is “can you please explain this to me like I was a 5th grader” and for that kind of thing AI will be revolutionary to productivity (like having a calculator that adds big numbers together). Humans can make a mess of explaining when they don’t have to, and take a lot of time doing it. But the messiness is also where the good stuff, the random stuff, the fun and exciting stuff comes from—the entertainment, the art—and for that, AI is not going to suddenly look at every book ever written and happen upon a eureka moment. For computers, the creative leap from pre-teen to high school to Stephen King might take more than a few gap years.
Thursday Writing Advice: Garbage In, Garbage Out
The other big talking point about AI for creative people is its possibility to be used as “a tool.” Ann Lamont, queen of writing advice, popularized the idea of the “shitty first draft,” and a common opinion about AI is that it can serve as a good shortcut to a shitty first draft. The reason I would caution any creative person using AI as “a tool” for actually doing creative work and not just for research or a fun lark, is that while you raise the floor of the time it takes to write a first draft, you also put a ceiling on creativity. Editing is not writing. Tinkering endlessly is not the way to a masterpiece and starting with a generic, logical but lifeless, machine-generated chapter is most likely going to start you on a bland, average path. Best-case scenario, it will spark you to re-write it all in probably the same time you could have done it from scratch.
In manuscripts of all varieties and types, I see this happen on a grand scale. You’ve probably seen it too in your own reading. Once one bad choice is made by a writer—a stray plot point, a shaky sentence, a flimsy source—you can almost guarantee other missteps will be made down the road and compound the problem. The saying that always pops in my mind to categorize this: garbage in, garbage out (ironically a phrase that was popularized in the early days of computers). When you start to see errors or bad choices crop up in a book, it almost ensures that other problems are on the horizon. If you want to write stories or arguments or book descriptions that have worth, then do yourself a favor and don’t start with garbage.
A Page at a Time: Moby-Dick by Herman Melville, or The Reading of a Daunting Classic
Progress: Days 118, Pages 557
Time to gloat. A few weeks ago, I predicted that there would be a giant storm before our time with Moby-Dick was over—being that there are only so many things plot-wise that can happen on a boat—and now, I can report to you, this has come to pass. As we near the final nautical miles of our journey, the story and characters have come back into focus. Like Nantucket and getting the crew together at the start of the novel, we’ve got a more traditional narrative as we build toward the final chase. One could view the shape of Moby-Dick as a classic, flawed “sagging middle” kind of a novel with all of the action coming at the start and the finish with lots of meandering in the middle. However, there’s a strong case that the wandering may be the point and we, the reader, are Ahab wading through arcana, seeking an elusive story. With only two weeks to go, I haven’t decided which camp to strike up residency in. But either way I am beginning o form plans to put the sagging middle theory to the test, stay alert, be on the lookout for a new kind of the whale.
Meanwhile, one of my favorite chapters in the book so far, which is easy enough to reproduce here by hand in its entirety goes as follows (no spoilers):
Chapter 122
Midnight Aloft—Thunder and Lightning
(The Main-top-sail yard.—Tashtego passing new lashings around it.)
“Um, um, um. Stop that thunder! Plenty too much thunder up here. What’s the use of thunder? Um, um, um. We don’t want thunder; we want rum; give us a glass of rum. Um, um, um!”