Note:

This excerpt is part of a longer, unpublished project on language and artificial intelligence.

“Fine,” my dad said. “Take the medicine if you want. If you don’t want to take it, that’s fine, too!”

We three were in the kitchen of our little house in Palo Alto, California. My mom poured the thick amoxicillin suspension into the plastic thimble of a medicine cup. I was not yet two years old and it was my ear infection we gathered in the kitchen to treat. This was not the first dose, so we all knew what we were in for. My dad, a six-foot three Air Force officer and weightlifter held me with one arm and picked up the cup with the other. My mom, she was also in her mid-20s, is a diminutive woman, especially next to my dad. She prepared for what she knew would be a struggle to hold my arms down.

I don’t know if you’ve ever tried to hold a child of not-yet-two who doesn’t want to be held. I once held my own son while a doctor and nurse tried—tried—to draw blood from his arm. Kids that age have a spastic, writhing technique that cannot be taught and a core strength that borders on super-human.

In one corner were two grown adults. And in the other, a toddler on a mission. As the sun set over the Stanford campus, we were ready to rumble...

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

I was at Google’s headquarters (The “Googleplex”) in Mountain View, California last month for the Imagination in Action conference. There is plenty I could write about—a Distinguished Researcher at DeepMind who talked about how we’ll learn to use agents over time; a leader from the non-profit sector saying that she’s done hearing pitches from companies that wrapped a piece of software around an LLM to make it “non-profit-friendly” asking for $10 million; or the tech CEO who, when asked about mentoring said, “if you want to get promoted, do real work.”

But I want to focus on a very brief comment—almost a throwaway line—I heard from a leader in Google’s Gemini program. He talked about how he’s using Gemini as an “ambient companion” while doing things like playing video games. Gemini shadows him as he plays. When he gets to a place in the game where he has to perform repetitive tasks, he can say, “Gemini, would you take over here and level-grind for a bit?”

Gemini is just sort of there, alongside him, prepared to takeover when he gets bored, or disinterested, or a higher priority task comes up.

This line is the kicker. He said, “I can’t tell yet if it’s important or creepy, but I’m playing around with it.”

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

When I was young, I used to build those plastic model airplanes you can find at hobby stores. When they were finished, instead of putting them on a shelf, my dad helped me to hang them from the ceiling; so, it looked like a dogfight between an A-10 and a P-47 in the corner of my room.

One morning, I woke up to the sound of my dad’s voice saying, “Joey, Joey! Get up! There’s an earthquake!”

Los Angeles and its San Andreas fault are famous for earthquakes. We lived through several that were 5-plus on the Richter scale.

I sat up, rubbing sleep from my eyes and, with all the wisdom of a six-year-old, asked my dad, “how do you know there’s an earthquake?”

“Look at your models,” my dad said. I glanced at the ceiling in the corner of the room. Sure enough, what had been a static snapshot of a dogfight had become bouncing and swaying model aircraft—both submitting to the motion of the quake and defying all the principles of flight.

I decided that evidence was sufficient and crawled down the ladder from the top bunk to join my dad and my sister under the door frame.

From that moment—the moment I, at six years old, asked my dad, “how do you know there’s an earthquake”—I’m sure he knew I was doomed to a life studying philosophy.

I attended a philosophy conference—a real philosophy conference—last week...

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

I couldn’t sign a PDF. I don’t know why I couldn’t sign it. I could sign other PDFs, just not this one. I finally joined a Teams call with an IT technician. He took control of my computer and checked and rechecked settings. Everything looked like it was set up properly. He couldn’t figure it out either. Ultimately, he told me to have the author of the document rebuild the PDF to see if a new version would work.

After we hung up, I received two emails. The first was the auto-generated email to say the tech had closed the trouble ticket.

The second email was from the technician. In it, he offered some amplifying information about the Adobe error. The text began with a quotation of the error messages I had received. Beneath that, in a bulleted and sub-bulleted list, were five steps I might consider taking to resolve the issue. I tried them. They didn’t work.

I had a hunch about what was happening, but I wasn’t sure yet. I have an officemate who uses generative AI tools more often (and more creatively) than I do. I called him over. “Would you take a look at this text? What do you think?”

He looked for about 30 seconds; first studying the substance to try to understand the error. But then he managed to look past the function to see the form and it hit him. “Oh. This is ChatGPT,” he said.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

When I was about 8 years old, my family and a bunch of other families from my church participated in a jog-a-thon. We asked friends and neighbors to support us by pledging a certain number of dollars per lap that would ultimately go to the designated charity. Then, on the big day, we kids would run as many laps as our little legs would carry us. Then, our friends and neighbors were on the hook to donate what they had pledged.

The track we ran on was probably a standard quarter-mile, oval track. Like most seven year olds, I hadn’t taken geometry yet. As I started another lap on the large arc that was the end of oval the track, I thought to myself, “why should I bother running around this curved part? It would be much easier just to run straight across the field. That would get me to the end of this curve faster. It would save me some time, it would probably enable me to run more laps, and that would raise more money for the charity.” It was a pretty brilliant plan...

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

My Grandpa used to do a magic trick with his dog, Halley. He’d tell gullible audiences (usually us kids) that Halley knew how to count. Then he’d hold out a milkbone and tell Halley, “count to eight.” Halley barked at the milkbone exactly eight times—then my Grandpa would gave the dog the bone. She stopped counting at eight… magic.

I enjoyed this trick several (ok, many) times before I realized that the dog would have kept on barking to infinity. It was my Grandpa’s act of giving Halley the bone that got her to stop barking at the right time. In other words, the dog wasn’t counting to eight. My Grandpa was.

Large language models are a little like Halley, the dog. ...

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

The global AI phenomenon we are all living through began in 2022 when OpenAI launched ChatGPT, a chatbot based on a large language model called GPT-3. Within months, ChatGPT had gained 100 million users, making it the fastest-adopted technology in human history.

But why did ChatGPT see such wide adoption so fast? What is it about ChatGPT that resonated so strongly with us human users? ChatGPT wasn’t the first AI tool. It wasn’t the first large language model. It wasn’t even the first chatbot. So, what made ChatGPT so special? …

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

When I was 22 years old, I fell in young love with a girl named Megan. I had packed all of my worldly possessions into the back of my used Toyota Rav4 and stopped in Pennsylvania both to pick Megan up for the epic road trip to my first assignment in Las Vegas and to meet her family.

I slept in her little brother’s room. My first morning in that new and unfamiliar place, I woke up to the sound of someone—someone very like Megan—bursting into the room where I slept and jumping into bed with me...

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

There’s a line in C. S. Lewis’s space trilogy that sticks with me. In Out of The Silent Planet, the first book in the series, the protagonist, Dr. Ransom, is speaking to the villain, Professor Weston. Weston is a world-renowned physicist who has solved human interplanetary travel. Weston as brought Ransom on the journey against his will. When Ransom awakes in the spaceship, he realize what has happened and asks Weston, “how have you done it?”

Weston responds condescendingly,

As to how we do it—I suppose you mean how the space-ship works—there’s no good your asking that. Unless you were one of the four or five real physicists now living you couldn’t understand: and if there were any chance of your understanding you certainly wouldn’t be told. If it makes you happy to repeat words that don’t mean anything—which is, in fact, what unscientific people want when they ask for an explanation—you may say we work by exploiting the less observed properties of solar radiation.

The line is brilliant in part because it helps to establish Weston’s megalomania, which plays an important role in the story later on. But it also raises a question about how much “unscientific” people can understand scientific things.

In my own profession, we tend to impose a distinction between “technical” people and “non-technical” people. I first arrived at the Pentagon just after finishing a PhD in philosophy.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe

I pushed the throttles up slightly, increasing power and I banked ever so slightly to the left and then back to the right. This put a little extra lateral space between me and the lead aircraft. As I passed him—another T-38 student pilot named Drew—I gave an exaggerated head nod, throwing my chin all the way down to my chest to make sure Drew could see it from his jet. With that head nod, I had taken the lead. I pulled the throttles back slightly to return to the original power setting and airspeed. I led the other aircraft to the high pattern above the Oklahoma farmland. Then I tapped the palm of my gloved hand—open, so he could see my fingers outspread—up against the canopy glass a few times. The hand motion looked almost as if I was pushing him away. And that’s what I did. As soon as Drew saw my hand tapping the glass, he banked slightly right and fell back, repositioning from the tight fingertip formation we used for the formation takeoff to the more flexible route formation.

You see, I had over-sped my aircraft’s flaps on takeoff and there were several procedural steps I needed to take in response to that mistake. And so far, Drew and I had done all of them without saying a word to one another.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit chapainsights.substack.com/subscribe