Hello Reader,
Now that Daedalia is complete after 43 weeks of routine writing, editing, recording, and publishing chapters here, I find myself in a bit of a postpartum trough.I’ve decided to take a little sabbatical from fiction to write about real things. This series of essays is part memoir, part rant, and part prayer for a future where we all come out okay. If you know me, you’ll know that I’m not a regular practitioner of any of these things, so please be patient. I think there’s something here worth sharing.
Peace & music,
Ben
It’s 1996, 2 AM and I’m sitting in front of a computer monitor the size of an industrial microwave. I’m watching with great anticipation as a JPG I just created in Photoshop 3.0 loads one row of pixels at a time over my dial-up connection. This is my first website. I’m at peak anxiety. I have $436 in my bank account, a new baby sleeping in the room next door, and a terrified, sleep-deprived wife lying wide awake in the other bedroom, worried that I’m looking at porn and wondering whether we’ll actually survive this life choice we’ve made.
I’m a 26-year-old musician who just released his first album, teaches guitar, and hosts an open mic night. I earn about $12,000 a year. With the help of one of my students, I’m teaching myself HTML. In a couple of months, he will hire me as a “web designer” at the American Cancer Society. It will mark the beginning of the end of my dream of being a professional artist. But I will be able to feed our baby.
For the next thirty years, I will get up and go to work through a series of jobs that are increasingly complex and demanding. My ambivalence about the ladder (a childhood obsession with Donkey Kong notwithstanding) means that some days I climb, and other days I look out the window. I’ve struck an unlikely alliance with the computer. It can do curious things I find fascinating, but more importantly, unlike my guitar, it can produce things people will consistently pay for.
If you work with computers, if you’ve programmed them, you understand what three decades inside the machine looks like. It’s toil and tedium. One step forward and two steps back, then a game-changing turn with unexpected consequences, but always iterating, always progressing. This was business as usual in my field until generative AI completely changed the entire fucking board and all the pieces on it.
Today, I spend 85% of my workday literally talking with my computer, specifically a variety of large language models (LLMs). Claude, ChatGPT, Gemini — these are now my co-workers, and my survival in this industry depends on them. I must report my usage of AI in a spreadsheet every month to my employers to justify their investment in the technology and, I suspect, to keep an eye on when things tip in favor of less need for carbon-based workers. Combine this with the fact that AI is generating terabytes of slop novels and songs that make it even harder for my own novels and songs to find an audience, and you’d think I have a solid case to absolutely hate AI.
I don’t. I think it’s a technology with the potential to make more of a positive impact on humanity, society, and the systems we’ve built than any technology that’s come before. It will challenge everything we think we know about ourselves, our world, and our place in it. It can mirror, emulate, amplify, and expand the best ideas humans have brought to this planet.
Or the worst, which means it could kill us all.
So, what happens next? It’s really all we care about, right? What happens next is the simple question that’s kept us all alive. That baby I had no business having 30 years ago is thriving today because I paid attention to this question. I learned a story from my mother and father that there is no higher priority than taking care of your family, and I followed it. I trusted the story to be true, and it was. But at an elemental level, stories are nothing more than predictions of possible outcomes — training data. Not all stories can be trusted.
When I first encountered the power of generative AI, it was GPT-3 on OpenAI’s developer preview in 2020. I had just wrapped up my novel, The Memory of My Shadow, so I had been thinking deeply about this alien technology invading human domains for a number of years. But I was writing fiction. When I started chatting through this nondescript little interface and its responses, painted to the screen one line at a time, were lucid and even clever, I was gobsmacked. I couldn’t get enough of the magic trick. In the next two years, the models got exponentially better, and I began chatting with them about writing and sharing chapters from my novels for feedback and insights. How utterly amazing to have an intelligence that’s read every book in the world respond to my work, right?
The bloom quickly faded. The cheap seduction of the LLM’s sycophancy and the wild liberties they took with my prose were enough to open my eyes. Oh, right, this thing’s not thinking, it’s a big math model placing bets on words that most commonly show up next to each other. A card trick and nothing more. I moved on and didn’t pay attention unless there was a new announcement. Then AI showed up at my day job like the plucky, idiotic intern asking to be helpful but mostly getting in the way and causing a lot of rework. It was good for a laugh. It generated some crazy pictures and I went down a rabbit hole with those for a while. It was all Cheez Whiz and Nutter Butters — empty calories. Then it got serious.
The LLMs coming out of the so-called frontier labs (how you call multibillion-dollar enterprises “labs” is quite a trick) were leveling up every month and showing they could do real work. I began to tinker more, to push the limits of what the models could do, and soon the tinkering was no longer a diversion, but a measurable requirement for my job. In every company of a certain size around the world, employees performing “knowledge work” are being asked to track specifically how much AI is improving our efficiency or, put more plainly, hoisting the guillotine for our future dates with permanent retirement. I’ve been laid off twice in my thirty-year career, and the sting of it has never left me. So, you see, I am the proverbial frog in the pot. I never set out to do this thing. It was a means to feed my family and a way to pay for the privilege of making music and writing. And now, it’s too late to reinvent myself. The only sensible thing is to double down and gain fluency with AI. It’s just another technology, right?
After a couple of decades of mostly leaving my work at work, I began blurring the boundary just as I had at 26 when I was staying up until 2 AM learning how to write JavaScript. Back then I was desperate to take care of my baby. All these years later, that calculus hasn’t changed much, only this time, it’s a different child — a brilliant, neurodivergent, grown child who, like so many of her generation who have graduated from great universities with honors is struggling to find her footing in a world that’s on fire, full of broken systems, and an ever-shrinking middle class. So, more than just digging a moat to protect my career, I decided I needed to start building a lifeboat for her in the form of a software startup, ironically using this tool that she and most of her peers fear and loathe.
This is how I end up in my underwear standing at the kitchen counter with my laptop far past when I should be in bed. I’ve been furiously building something for months, speed-running through the full development cycle that, until a year ago, would have taken a team of six people twice as long. I’ve given it the time and dedication I’ve typically only devoted to my music and writing. It has become something of an obsession, which might be hard to understand or may seem distasteful or even treasonous, depending on the zeal of your stance on AI. You might ask how someone who claims to care so deeply about human stories can spend hours with this word machine. A machine built by oligarchs, one that threatens to save or destroy humanity depending on the day and who you talk to. How can he trust it?
The honest answer is, I don’t fucking know. That’s why I’m writing this series of essays. I want to understand myself and how this technology is changing me and the society we live in. The truth is no one has the answers. There are plenty of people who will tell you otherwise, plenty of people entrenched in one of two camps. God, do we love a good binary — that’s one area where we have a lot in common with computers. But nothing about this is a pattern we’ve seen before in human history, so it’s impossible to make anything but the wildest predictions of where it will take us.
I’m doing this experiment on myself, grappling with this thing in real time, for a myriad of contradictory and confounding reasons. As I finish this essay, once again up past my bedtime, Claude is whirring away in another window on my laptop, running a battery of automated tests to validate version one of this product I’ll be shipping soon. Each line streams in with a green checkmark, and I’m filled with excitement and dread in equal measure. Is this story a happy ending? What is a happy ending in these times?
What happens next?
This is one of three essays in my Human in the Loop series. Next week, I’ll dig into the toll AI is taking on me, some unexpected upsides, and generally how I’m managing to navigate this brave new world. This is a wildly charged topic. If you have thoughts, dive into the comments.