Hills beyond the horizon

28 December 2020
28 Dec 2020

This is an excerpt from today’s issue of my weekly newsletter.

What makes a great conversation?

Certainly there’s some degree of give and take, listening to the voice of your interlocutor and responding in kind. You want to move fluidly between topics, avoid jumping from idea to random idea without transition. In the best of times, it feels like exploring a mental landscape that you share with your partner-in-conversation, mapping out unknown territory shared by the two of you and no one else.

But the best conversations I’ve had are often the ones punctuated by randomness and non sequiturs.

In artificial intelligence, there is a problem called the hill climbing problem. One way to frame artificial intelligence is as a search for efficient ways to optimize some goal value, so computer scientists sometimes model a goal “function” as a kind of a terrain, where you start someplace in space, and you explore the terrain to “climb” up hills until you find the tallest peak. (AI experts—forgive me for my oversimplification for illustrative purposes!)

A naive, easy solution is to simply take each step in the direction that increases your elevation. This leads you to one peak, but in all likelihood, it’s probably not going to be the tallest one. You get stuck in a local maximum, the tallest point in some neighborhood that ignores some taller peak farther away. It turns out better solutions involve some random chanced exploration. An approach called simulated annealing involves making random, obviously bad moves once in a while that might land you closer to a taller hill.

I think navigating a conversation sometimes feels like this—you’re trying to find pockets of common interests that can spur interesting dialogue and ideas, and if you only move from topic-to-related-topic, it’s easy to get stuck in mildly interesting “small hills”. Once in a while, you need some randomness, some total out-of-left-field interjection or idea to help you find more exciting ideas to jam on.

Navigating the terrain of a career can also be similar. Chris Dixon has written about this exact framework for navigating career paths.

I think creative work can also feel this way sometimes. If you want to create something new, you can’t just iterate from the first decent idea you have, you need to spend some time jumping around different ideas, accepting the chance that you’ll have some truly terrible ones, to explore the terrain fully before you pick the tallest hill to climb.

When we choose our next steps by only taking steps upwards, we keep ourselves blind to the complexity of the terrain that can lie before us, the troughs and hills that make life interesting. Exploration is about chances. If we keep ourselves open to the occasional game of chance or random jump to something new, we might stumble into even grander landscapes stretching out before us.

Looking back at my writing in 2020

Laying bricks, building castles: sketches aren't perfect