It seemed simple. I asked Siri to play They Might Be Giants—it turned on They Might Be Giants Radio.

“Can’t Keep Johnny Down”
“I Love You for Psychological Reasons”
“Circular Karate Chop”

I was impressed that it went last decade early in the mix.

As we drove up 127, slowing down before Uncle John’s Cider Mill, I was grooving on the tracks with which I was far less familiar.

Then, as we sped up at Ithaca, it dawned on me.

Aside from a couple old favorites, “Don’t Let’s Start” and “She’s An Angel”, I noticed it played only tracks from Phone Power, Nanobots, and Join Us.

I started skipping ahead. More tracks, all from those three albums.

Of course.

I’d already trained the AI.

We’d had Spotify for a year before I realized I could listen to music I’ve never heard before. The albums I’d always wanted to hear. So I went through some of TMBG’s albums I hadn’t kept up with. Namely those three albums Spotify played for me.

And it favored Join Us because it was the one I played more than the other two.

At the same time, the albums that shaped high school and college, Flood, Apollo 18, John Henry, and Factory Showroom were entirely absent.

It is time to train the AI.

I needed the app to know that Flood is the beast of them. So I put it on play and proceeded to like the essentials. Not just “Birdhouse in Your Soul”, “Istanbul (Not Constantinople)”, and “Particle Man”, but “Dead”, “We Want a Rock”, “Letterbox”, and “Whistling in the Dark”.

And when that was over, we moved in order.

The AI needed to know everything is on the table for TMBG Radio.

It doesn’t know what we know, it watches what we do.

I was surprised TMBG Radio wasn’t generic, playing curated hits. Of course, I should have known better. That’s the 2010 internet radio experience. Much like my Punk Radio is influenced by the tracks and bands I listen to, this was the stuff I’d previously played.

What surprised me was how it was entirely based on its experience of me. It doesn’t have a proverbial mind of its own.

When I played those certain things, it designs a playlist that includes only those certain things and maybe a smattering of other songs.

Even with a playlist of 302 songs, it centered on the ones it was trained to.

AI is designed to observe our behavior and seek to match it.

In this way, machine learning is a bit like human learning.

Like my kids learning how to exist in the world by watching how I exist in the world. When I hang back, they do too.

If we want our AI, our families, and even ourselves to be a different way, we have to make that way happen. Which literally means doing that different way.

Playing what we want to play. Now and later. Being bold. Generous. Hopeful. Silly.

Having fun. And making sure our playlist is just as fun.