The world is not a collection of billiard balls

against mechanism

I am grateful to Isaac Newton for having invented a theory which, for a time, gave us a deeper understanding of the cosmos than we’d ever had before. I am not grateful to him for indirectly causing many people, hundreds of years later, to take too seriously the picture of reality that his theories left us with.

What picture of the world do Newton’s laws evoke? Here I’m not interested in the specific equations of motion he wrote down, or his technical definitions of force, acceleration, and inertia. I am interested in the intuitions and metaphors underlying his theories. As humans, we understand everything by metaphor¹ to our basic, everyday experiences, and for most people, Newton’s laws draw on very particular metaphors:

The world is like a bunch of billiard balls. What are the salient properties of billiard balls from our perspective as humans? First, the balls are atomic: they are indivisible and featureless units. There is nothing too interesting about them beyond their interactions with other balls. Also, they are inert and unalive. They don’t feel anything of course, they’re just rocks.

And what about their movements and interactions? Their movements and interactions are all predetermined. Strike the balls in the same way, and they will follow the exact same paths. Also, their interactions are all local: the only thing influencing the movement of one ball is the direct contact it has with other balls (or with the edges of the table). Everything that happens on a billiard table is just a bunch of bumping and shoving. Out of that bumping and shoving comes everything we care about in a game of billiards, all the patterns and strategies and gameplay narratives. It’s a bunch of inert, uninteresting, featureless rocks bumping and shoving against each other.

The “scientifically minded” person who has not spent much time thinking through their metaphorical handles on reality sees all of these things as properties of the world in general. The world is a bunch of billiard balls that move. The world is a bunch of puzzle pieces or lego blocks. The world is fundamentally dead, predetermined, and unwittingly, imperturbably subject to the progression of math equations. What’s interesting is that even after realizing that the theory itself is technically untrue (quantum mechanics has superseded Newton’s laws), the metaphors underlying the theory stay with us.

There is a philosophical name for the worldview that results from this picture: mechanism. And mechanism, as a way of looking at the world, is one of the most powerful things we have ever come up with. Thinking of the world in terms of local bumping and shoving, featureless subunits that grind against each other like gears and pulleys, has given us, well, incredibly useful inventions. And it continues to be a useful orientation for science today: we seek a mechanistic understanding of black box AI systems, and we try to decode the marvels of life through mapping out its biomechanical foundations – the shapes of proteins, the graph of neurons in your head.

Of course, we also know today that mechanism is not the full picture of reality. The elementary particles of reality are not actually featureless (and they might not be unfeeling), local “bumping and shoving” is not the only legitimate form of interaction, and the universe might not actually be fully determined by its initial conditions. Research programs that operate exclusively in terms of mechanism have often run into dead ends, like with the recent slowdowns in progress in mechanistic interpretability, or the failures of reductionism in biology.

But there is something else that makes the mechanistic picture especially hard to let go of: the possibility of control. That is the promise that mechanism makes: once you truly have a mechanistic understanding of something, you can interfere with it, disrupt it, reshape it to fit your goals. This is why AI safety researchers care to have a mechanistic understanding of AI systems, so that we can maintain control over them, lest they become too powerful.

To let go of mechanism, then, is to accept that there is a limit to your capacity to control the world. The mechanist believes that reality is a mechanism all the way down, which means that reality can be controlled all the way down. The more we progress in science and technology, the more fully we can control reality to fit exactly to our liking.

I don’t think we can entirely settle this question today, with philosophy – history is replete with both examples of our failure to shape reality to our liking (see every dictator that was toppled, the long list of diseases we have yet to cure, and the fact that we still can’t get computers to work right), but it is also full of feats of control we would not have imagined possible (gene editing, moon landings, and short films made out of individual atoms). We may just have to let the mechanists have their chance at building more technology and see how powerful it makes us. But there is another question I want to leave you with, which might be easier to answer today. What kind of person do you become, when mechanism and control are your operative lens for viewing the world? Is that the kind of person you want to be? And is that the kind of person you want to share a world with?

1. See Metaphors We Live By, or Surfaces and Essences for more on what I mean. For a brief intro check out my tweet thread or this article.