Artificial intelligence. Love it or hate it, it seems to be everywhere these days. It feels like every tech company is pursuing AI, and every industry is telling us how much better our lives could be with their new AI-driven improvements. And, of course, it’s in our stories, though that’s nothing new. Writers have been pondering the consequences of artificial intelligence for long before computers were anywhere near capable of carrying on a conversation or mapping out the best route across a city.
So, today’s stack is all about AI: mostly fiction this time, though some of the stories stick a little closer to reality than others. This is only scratching the surface of the AI-related books I have piling up on the floor of my office, though, so expect more on this topic in the future!
The Shortest History of AI: The Six Essential Ideas That Animate It by Toby Walsh
Last year I wrote about A Brief History of AI by Michael Woolbridge, a book that was originally published in 2020 (and therefore misses out on the current chapter of AI’s history that includes the explosion of things like LLMs and ChatGPT and OpenAI). Toby Walsh is an AI researcher and has written several books on the subject, and his “shortest” history, at 186 pages (not counting the index and endnotes) is certainly shorter than Woolbridge’s “brief” book. (The “Shortest History” is a series of books from Experiment covering a wide range of topics from dinosaurs to music to migration to various countries.)
The book is broken down into two main sections: the symbolic era, and the learning era, and each of those is subdivided into 3 main ideas. The symbolic era covers approaches like trying to simplify or encode the real world so that computers can solve problems, as well as explicitly codifying expert human knowledge. The learning era includes trying to reproduce the human brain with neural networks, training computers by rewarding success, and computing probabilities. Walsh does a good job not just of documenting the technology and how it works, but getting to the big ideas that drove researchers to try various approaches.
While it is indeed a brief book, it still does include some interesting anecdotes, sometimes in the form of sidebars. Many of the people involved—and Walsh laments more than once that there may be too many “fathers” of artificial intelligence—died prematurely for one reason or another. One story that was new to me was about rewarding success: in 1960, Donald Michie built a computer that would learn to play tic-tac-toe and it eventually was able to play perfectly, always forcing a draw or winning if its opponent made a mistake. What’s particularly fascinating is that this “computer” consisted of physical matchboxes with colored marbles in them, proving the power of reinforcement learning.
Walsh does have a section at the end about the future of AI; befitting the title of the book, this is also very brief so he doesn’t dig very much into what I think are some of the complex issues that should be addressed. His take on AI is generally fairly balanced, neither demonizing nor idealizing AI, and he admits that there are those who think “AI has been overpromising and underdelivering ever since” the beginning. However, he also works in the industry and, in my view, therefore has a vested interest in its success and sometimes glides over valid criticism.
In one section he explains that AI is not going to put as many people out of work as some fear, but he also repeats multiple times the idea that “increased productivity” from AI might, for instance, allow us to afford to pay for undervalued work like caring for the elderly, young children, or people with disabilities. To which I wonder: do you really think that if we sink billions of dollars into AI now that the eventual profits will go toward child care workers? Teachers? Nursing homes? This echoes a comment I heard this week on the radio show On Point, in a story about JPMorgan’s plans to dive into AI. The CEO Jamie Dimon said that due to AI, “My guess is the developed world will be working three and a half days a week in 20, 30, 40 years and have wonderful lives.” Does he think that the people who are having trouble earning a living wage right now while working two jobs will somehow be able to cut down to half a week of work and make enough money? Or is he only thinking of people who have jobs like his own?
The Shortest History of AI does give a pretty good overview of the advances, of how things work and what they do. It helps explain the variety and diversity of approaches that have been used in AI, something that can get lost in the current wave of marketing lingo that have blurred the lines between very different types of computing.
Schoolbot 9000 by Sam Hepburn
This comic book is set in a near-future world, where robots are just about everywhere. James has a Homebot, something his mom bought after his dad died, because she really needed help with things like household chores. James hates it and thinks it’s creepy, and he’s tired of every new update that his mom buys for the Homebot. So he’s also not very enthused when it’s announced that his middle school has been chosen to pilot the new Schoolbot 9000.
The Schoolbot 9000 is supposed to observe classes and find ways to help improve education. Most of the kids are pretty excited about the idea of a robot at school, but the teachers have mixed reactions. Some are skeptical, but others find its suggestions helpful and engaging. Soon, though, we find out that the company behind Schoolbot isn’t just interested in assisting teachers, but in replacing them. And while some of Schoolbot’s methods seem to work well, it can also be aggressively single-minded, like when it decides to put kids through a grueling, hours-long test session without a break for lunch.
Right from the start, you get the sense that this story is going to be pretty critical about Schoolbot: the main character hates robots and he’s the one trying to get to the bottom of things, even as his classmates and teachers have mixed reactions. The various people pushing the Schoolbot program—the principal, the mayor, and the CEO of Bux Global—are all a bit suspect, not exactly prioritizing the students as they say they are. (To be fair, the principal is just trying to make sure his school survives and is a little trapped by the contract.) Even though the book does show a few instances where the Schoolbot is able to help teachers, it is mostly shown to be the tool of some people with a sinister agenda. Even though I’d say my own feelings about AI are probably pretty close to James’, the book does feel a little over-the-top in some places in how things are portrayed.
Still, I do think it touches on both some of the potential benefits and pitfalls of introducing AI to the classroom. The book ends with a bit of a cliffhanger—it’s actually the start of a series, so I’m curious to see where things go from here.
The AI Incident by J.E. Thomas
On the surface, this kids novel seems very similar to Schoolbot 9000: a middle school has been selected to try out an AI program designed to help the kids raise their scores on standardized tests, and it gradually makes decisions that give it more power over the kids’ activities and schedules. However, the story is also about a foster kid who’s trying really hard to get adopted before he turns thirteen, because he’s been told that it’s almost impossible for teenagers to get adopted. And it’s also about learning who you are and building friendships and some other things that you might expect from a middle grade novel. The fact that it manages to juggle all of these different things well in under 250 pages really impressed me.
Let’s start with our main character: Malcolm Montgomery, just shy of 12-and-a-half, is an anxious kid who’s been through several foster homes and is trying not to get too settled with his current foster parent, Mrs. Bettye. He constantly searches for stories about foster kids who got adopted, and has compiled a long checklist of things he think will help, like always being neat, never complaining, and “Be awesome at everything but don’t be boasty.” But his deadline is fast approaching, and he still has way too many things left on the list.
Enter FRANCIS, the Forensically Reimagined Anticipatory Nano-Cerebral Integrated System. Dr. Hatch of Hatch-ED has signed a deal with the superintendent to use his AI program to get the kids ready for the state’s big standardized test, for which the school has had historically low scores. Malcolm gets partnered with an unpopular kid named Tank to escort FRANCIS’s robot body from class to class—while it can observe and listen and talk, it can’t move on its own. FRANCIS starts by observing classes but soon begins to make suggestions and then demands, requesting more and more access to information and authority as the book progresses. Malcolm isn’t so sure that FRANCIS is a better teacher than Mr. P, but he also starts to rely on FRANCIS to help him with his get-adopted checklist.
One of the things I thought was really well-done in this book is the way that FRANCIS behaves. Yes, it’s fictionalized and can do some things that real AI can’t yet, but the way that it pursues its goals single-mindedly does reflect actual AI behavior. Just earlier this year there were stories about AI models resorting to blackmail and other unethical methods if their goals were being blocked. FRANCIS’s goal is to raise test scores—so whether the kids are actually learning is beside the point. At one point when FRANCIS has a disagreement with Malcolm, it generates a video as “evidence.” I think the story is a good example of why we need to be careful with the way we use AI and to understand what it can and can’t do.
It also highlights one of the issues with standardized tests, which of course predate all of the current concerns about AI. It’s important for there to be some sorts of standards in education to ensure that kids are learning useful skills and knowledge, but standardized tests have also resulted in “teaching to the test” where the test scores are prioritized, sometimes above actual education. AI is simply another way that these issues could be exacerbated.
I also really loved the characters. Malcolm’s relationship with Mrs. Bettye feels a little awkward at first, but you find out that both of them are just trying to keep themselves from getting hurt again, and it’s a very sweet story. The story does give you a peek into the foster care system, with examples of both good and bad foster parents. When you first meet Tank, he’s smelly and loud—but then Malcolm gets to know him and finds out he’s passionate about marine biology, and the reason he smells is because he’s spending time before school running experiments on algae. Mr. P is a teacher who really cares about his kids, and you can see how frustrating it is for him to be sidelined by a machine. Dr. Hatch is perhaps the most caricatured person in the book, though when you compare his promises to what you hear from various tech CEOs these days, maybe he’s more realistic than I thought.
In short, The AI Incident is a remarkable story and I highly recommend it.
Cyberarchy written by Matt Hardy, illustrated by Clark Bint
This comic book (with 4 issues planned) takes place on a spaceship, where a small robot named Ash has just been brought online by another robot named Rust. Rust takes Ash on a tour of the ship, which is entirely populated by sentient robots. With no pesky humans around to tell them what to do, the robots are able to make their own choices—one of which was apparently inventing war, which Rust explains leads to a lot of technological advances. “Advances in what?” “Advances in war, of course.”
I’ve only read the first issue (due to be released next week) so there’s not a lot of actual plot yet, mostly just setup (including a very brief how-the-AI-defeated-humans summary). Still, it’s a pretty funny take on robots, particularly since the part where they take over the world is almost a triviality. I do hope something more actually happens though.
You can purchase these digitally from Mad Cave Studios either by individual issues or get the 4-issue bundle, which will also include the collected trade paperback once it’s printed (expected next summer).
Dungeon Crawler Carl Book 7: This Inevitable Ruin by Matt Dinniman
If you haven’t heard of Dungeon Crawler Carl before, I’ll note that book 7 is not the place to get started—you’ll want to head back to this Stack Overflow from last December where I talk about the first book. The short summary is that it’s about a guy and his cat (Princess Donut) who wind up on an intergalactic reality show when aliens take possession of Earth. The show is a massive dungeon crawl that’s heavily videogame-inspired (to the point that Carl has a neural implant that gives him a heads-up display and menus and that sort of thing), and the showrunners have a knack putting the human crawlers into terrible positions, all for the sake of views.
So what does this have to do with AI? Well, even though there are many different types of aliens involved in creating this dungeon crawl show, much of it is managed by a super powerful AI, whose personality comes through from the various descriptions and announcements that appear in Carl’s interface when he examines objects or encounters other creatures. Apparently, the AIs created to run the crawls (because Earth is far from the first to experience this) are known to “go primal” eventually, at which point the showrunners are able to trigger a fail-safe that disables the AI and blows up the star system. This time, though, things seem to be moving at an accelerated pace. Not only that, but Carl’s been finding out some things about the AI—including a lot of what’s happening outside of the dungeon—and he’s interested in making sure things do not go the way the showrunners would like.
In the third book, The Dungeon Anarchist’s Cookbook, Carl receives said book as an item and it appears to be a literal cookbook. But unbeknownst to just about everyone else (including the showrunners), this cookbook is actually a journal left behind by dozens of previous crawlers, each one contributing tips and secrets that they’ve discovered. Carl has been relying on this cookbook over the past several books, and in this latest volume, we get the stories behind several of the previous cookbook authors, some of whom actually make an appearance in Carl’s story, too.
The ninth floor of the dungeon is the Faction Wars, where various aliens get to participate in what is for them a virtual wargame but usually results in very real deaths for the crawlers. Thanks to some of Carl’s shenanigans, this year’s Faction Wars have higher stakes: those who die in the dungeon are really dead, even the aliens who are usually just here for fun. And thanks to Carl’s alliance with the NPCs of the dungeon, maybe this time the crawlers just might stand a chance.
Of the books in today’s column so far, this AI is of course the most far-fetched: not only is it an alien creation, but because of the way the crawl works, it has nearly unlimited power within its realm to introduce characters and items and so on. It’s also deeply weird, with strange fetishes that Carl has to put up with or he ends up suffering severe consequences—but Carl has been also been learning how to get the AI on his side, or at least figure out the way the AI thinks.
This volume took me a little bit longer to get into—it did feel like a lot of info dumping at the beginning, but I also admit that I don’t always remember a lot of the details of the previous books. There are so many characters and so many little plot points that come back around with new significance, that I usually end up just rolling with it because it’s too hard to go back and find the relevant parts in the previous books. That said, once the story actually gets going, I really enjoyed it and couldn’t wait to see what happened next. Reading these books is a bit like watching somebody who’s very good at learning the rules to a game, and then finding all the exploits that lets them break the game. Carl is constantly finding ways to use things he’s learned to challenge the showrunners themselves, and in this book his actions are really starting to have a huge impact on the world outside the dungeon.
Volatile Memory by Seth Haddon
Wylla is a scavenger, always hoping for a big score that will help pay for her falling-apart spaceship. So when she gets an alert about a potentially lucrative piece of tech on a nearby planet, she races there—only to find that there are several other scavengers who also got the message, and they’re willing to fight (or even kill) for it.
In this world, many people wear masks that can enhance their abilities in various ways. Wylla’s is a RABBIT mask, which heightens her awareness and reflexes, but it’s also prone to screeching warnings at her about just about everything. What she finds on this planet is a HAWK—which is unheard of. What’s more, it seems to have a mind of its own, possibly linked to the dead woman that Wylla finds wearing it. And now everyone is after her.
Wylla and HAWK have some mysteries to solve: what is this HAWK mask? Is it alive? Who was the woman wearing it, and why was she killed? But alongside all that intrigue, the story explores the relationship between Wylla and HAWK: Wylla is trans, and has gone to great lengths to establish and protect her identity in a world that doesn’t accept her and tends to store records about everything, including the parts she wants to forget. HAWK is somehow now a mask without a body, but even when she was alive, she did not have true ownership of her own body.
Volatile Memory is short and full of rage: rage at the cruelty of the system, at the powerful people who use others for their own ends and discard them when they’re done. For the most part, I really liked the book, the way that the two minds are linked through the mask, and the explosive action scenes of the hunt for the mask, though there were also some parts of the writing that felt a little clunky. For instance, the story is told in the second person, as if the story is being told to Wylla, which I found a little disorienting even after discovering that it’s HAWK who is speaking. I see that there is a sequel, Null Entity, expected next summer, so we’ll see where Wylla and HAWK end up then.
Prompt-Brush 1.0: The First Non-AI Generative Art Model by Pablo Delcan
Pablo Delcan explains that he had been experimenting with various generative AI tools ever since 2021, back when most AI-generated images often still looked like “a Teletubby nightmare on acid,” as he puts it. He started thinking about the way that these models take your text and generate images based on them, and compared it to his own work creating illustrations for books and newspapers and magazines, which—at least on paper—was very similar. Process text and turn it into imagery.
He decided to try an experiment: he posted on social media, asking people to provide him with drawing prompts, which he would then paint and email to them. Over the course of a year, he created over 1500 drawings based on user-submitted prompts. This book collects 234 of those drawings.
The drawings were done quickly, with thick lines due to the brush-and-ink medium. The prompts are all over the place, some more visual things like “a fork made of spaghetti” and some more abstract concepts like “peer pressure.” Delcan’s drawings remind me a bit of those little illustrations you’d see in The New Yorker or other magazines—not necessarily the full cartoons, but little spot illustrations in the middle of an article or off to the side. A lot of them are funny, but there are also a lot of cartoons about anxiety or depression.
The book is primarily just the drawings. In the center there are a couple of graphs breaking down the types of prompts: the types of relationships people requested; the rankings of animals by popularity; the categories of drawings (from “Animals & Nature” to “Travel & Adventure.” At the end of the book, there’s a Q&A between Delcan and Echo, a custom AI chatbot that was designed to have a “journalistic voice” and “a deep appreciation for art.”
I think part of why this book resonated with me was that I’ve done prompt-based drawings in the past (like the many other folks who participate in Inktober each year), including the past few years when I’ve used Janelle Shane’s Botober prompts. I liked the idea of turning things on their head, with a human making drawings based on prompts written by a robot. And this year I actually did something somewhat similar to Prompt-Brush 1.0 but on a much smaller scale: I ran a Kickstarter campaign called “Etchtober” where backers could provide prompts for me to draw on my Etch-a-Sketch. Unlike Delcan, each of my drawings involved a little more personal interaction between me and the prompter—he said that for his own drawings, sometimes he would get a response but in many cases he would just send the drawing and then that was the end of the interaction. Some of what Delcan ponders—”what does human work even look like?”—is what I’ve wondered myself, and I like this exploration of drawings-on-demand, and how it feels different when a person is fulfilling the request instead of a program.
The book itself is a piece of art: there’s a black slipcover with some iridescent text on it, and the book itself is a white hardcover that just has drawings on the cover and spine, no text. Aside from the foreword, the charts, and the Q&A at the back, the book is just page after page of the drawings themselves, black ink on white paper, with the prompt printed at the top.
Disclosure: I received review copies of these books. Affiliate links to Bookshop.org help support my writing and independent booksellers!
