Westworld is the Buddhist Game of Zendo

“Master, does this murderous artificial intelligence possess the Buddha nature?”

“They’re always trying to error correct, make themselves more human. When they talk to each other it’s a way of practicing.” — Bernard Lowe

There’s a game I played once, many years ago, called Zendo.

It’s named after a Japanese meditation hall, though I discovered it in a coffee shop, on a bookshelf, among the boxes of Candyland and Connect Four.

The game pieces themselves are chintzy, and there are only two types: translucent, stackable pyramids of various sizes and colors, and a few black and white stones. There is no game board. There are no dice. It works like this:

One player (the “Master”) creates a secret rule for structures (“kōan”) to follow, and the other players (the “Students”) try to discover it by building and studying various kōans which follow or break the rule. The first student to correctly state the rule wins.

The Master, for example, could invent a secret rule like “a kōan has the Buddha-nature if and only if it contains a single yellow piece.” The Master then builds two kōans: one which exhibits the rule, and one which does not. The Master places a white stone beside the kōan which exhibits the rule. Beside the kōan which does not, he places the black. Then the students begin to guess by building their own.

“Master,” you might say, kōan in place, “does a kōan exhibit Buddha nature if and only if it has a single yellow piece?” If that’s correct, you’re no longer the Student. You’re the Master, and you make a new rule.

So he who guesses best, wins. But there are no losers. Every player benefits by observing play and seeing who wins (or doesn’t) when they correctly guess (or don’t) the secret rule.

It’s a simple game, but infinite. When you begin, rules may involve single qualities. The color of a piece, the height of a kōan.

Keep playing, and rules naturally get more complicated: color, size, orientation, pointing, touching, stacking, nesting. Kōans become towers. The question of the buddha nature becomes baroque.

It’s worth noting there’s a whole biscuit shooter full of analogous processes in the real, beyond-the-Zendo world.

In science, all theories begin as if you were building a Zendo kōan — that is, making a guess, inside your mind, about how the observed world works. The same process occurs in our day-to-day lives; we’re all just making educated guesses about the best way to live.

The game of Zendo

Or, like Angela the Host said in episode two: “Everything is bespoke. All you have to do is make decisions.”

“If you can’t tell the difference, what does it matter?” — an epistemological double entendre from Angela the Host

Kōan-building, too, is a way to describe machine learning; algorithms develop themselves by evaluating sample inputs and constructing a model.

With all types of kōan-building, human or machine, the further you travel from your beginnings, the higher the abstraction gets and the more complicated the patterns become.

Or to put it another way: Sweetwater, in the center of the park, is easy. You ride out of town? That’s when the real demented shit begins.

“I never understood why they paired some of you off. Seems cruel.”
— The Man in Black

So the game of Zendo, you could say, is a way of interrogating the world. You look at two kōans, judge the difference, and guess the kōan. This specific form of interrogation, it turns out, is an example of a Bongard problem.

Bongards, which are almost as fun as they sound, are a kind of puzzle. They were invented by the Russian computer scientist Mikhail Moiseevich Bongard, who was very much interested in the concept of pattern recognition.

Bongards, like Zendo, start simple. If you’ve ever taken an IQ test, you’ve likely seen one. The basic idea: you’re presented with two sets of relatively simple diagrams, say A and B. All the diagrams from set A have some common factor. That common factor is lacking in all the diagrams of set B. Your task is to find the common factor.

Here’s a simple one:

An easy Bongard problem

And this one’s harder:

A medium Bongard problem

If you’ve heard of Bongards, it’s probably thanks to Douglas Hofstadter, who popularized them in Gödel, Escher, Bach, his Pulitzer-winning 1979 page turner on how cognition emerges from hidden neurological mechanisms.

Bongards, Hoftstadter would say years later, greatly contributed to his understanding of cognitive models and artificial intelligence.

Harry Foundalis and daughter. Harry, a researcher in Hofstadter’s group, created a computer program, called Phaeaco, which could solve Bongard problems. He then quit his research because he felt intelligent machines would destroy humanity.

Think about it. To solve a Bongard, you’ve got to bounce back and forth between diagrams. Sometimes you remain within a single set. Sometimes you compare across sets. To discern rules, you have to do guesswork. “Perhaps shapes count,” Hofstadter wrote, “but not colors — or vice versa. Perhaps orientations count, but not sizes — or vice versa. Perhaps curvature or its lack counts, but not location inside the box — or vice versa. Perhaps numbers of objects but not their types matter — or vice versa.”

Even when one’s first hunch turns out wrong, it often takes but a minor “tweak” of it in order to find the proper aspects on which to focus. In other words, there is a subtle sense in which people are often “close to right” even when they are wrong. All of these kinds of high-level mental activities are what “seeing” the various diagrams in a Bongard problem — a pattern-recognition activity — involves.

This activity, Hofstadter argues, forms a core part of artificial intelligence. To discern important features you must filter out the superfluous or superficial and create analogies at an abstract level.

That is, you don’t just lump things into categories. You create new categories, and find resemblances and differences at that arbitrary level. Perception, in other words, involves intuitive guesswork and subtle judgments. It’s not just computer science. It’s what Dr. Robert Ford, peering down into Westworld, might call “art”.

Westworld, of course, also has two sets.

“This place is one thing to the guests, another thing to the shareholders, and something completely different to management.” — Theresa Cullen, explaining the levels of the Bongard problem.

The Guests in Set A all have some common factor or attribute. We assume that attribute is “humanness”. The Hosts in Set B, despite looking very similar, don’t share that attribute. We assume they’re robots. But the differences are hard to suss.

Inside Westworld, the Guests have difficulty telling themselves from the Hosts. “Having trouble telling who’s who?” Logan asks Billy in episode two, before leveling a gun at nearby diners.

The Hosts, too, have trouble telling the difference. They’re convinced they’re human. It’s only the Managers, looking down, panning their tablets from one set to the other, who can tell who’s who.

Although we, the viewers, looking from a further remove, begin to wonder: is there any difference between the two sets at all?

Westworld is a Bongard problem.

That goes for the theme park and the show.

“I knew that, in the course of solving a Zendo problem, players would need to see many examples of both kinds of koans, and that good players, in the process of testing their theories, would often build koans that they expect to be black. It seemed arbitrary to penalize one kind of play and reward the other. Indeed, I knew that “black” and “white” could just as easily be replaced by “purple” and “orange” without changing the game at all. I wanted to come up with some reward mechanism that didn’t favor black or white.” — Kory Heath

In Zendo, a black stone marks the kōan that doesn’t exhibit the Master’s rule; white, the one that does. It’s tempting to look at Westworld in that dualism.

Dolores Abernathy wears white — must be good!

The Man in Black wears the color of his name — all beer and skittles, that sumbitch.

And of course, just before Billy enters Westworld, he’s given the most Klosterman of choices: do you wear the white hat, or the black?

But in eastern philosophy, the concepts of black/white and good/evil are interdependent. So, too, in Westworld — both outside the park, and within.

The gunslinger Teddy Flood begins his train ride bathed in white sunlight, but thereafter wears the black hat. Bernard Lowe, head of programming, wears a white shirt in bed with Theresa the head of ops — but wears all black when questioning Dolores in the office.

These colors don’t represent a set alignment, in the D&D sense (Bernie is Chaotic Boring!), they represent lines of questioning.

See, for example, when a milk-loving host goes rogue in episode one. Bernie, wearing black and white, tells Theresa “It’s a good problem.” Bernie, like the hosts, is testing a theory. He’s interrogating the world.

This is the non-dualistic nature of the kōan. Don’t strive to see duality, see the singular nature of all. Ask yourself: What is the sound of one artificially intelligent hand clapping?

This is also the nuance between identity and agency. “They already know who they are,” says Dr. Ford in episode two. “They’re here for a glimpse of who they could be.”

In the process of testing that belief, they must build some kōans that they expect to be wrong.

— Dr. Robert Ford to Lee Sizemore, who did not guess the Buddha nature with his Odyssey at Red River kōan

In Westworld, it’s tough to tell who’s a player and who’s a pawn. Each person, whether guest or host, exists as a nested structure.

Is the maze a kōan in the shape of the human mind? Is the Man in Black a way of asking the question? Is Billy? Is either method fully right, or woefully wrong?

Or is Westworld itself the kōan—a colony of Vitruvian Man-ants that collectively create, in that Hofstadterian way, a functioning society based on rules? Is it all a way for management to ask a larger question?

It’s unknowns all the way down.

The kōan may get the white stone. The kōan may get the black.

But either way, the only certainty is that building a correct kōan, as in Zendo, only gets you one thing: more questions.