← Chetan Surpur

An Elusive Fly

There’s a fly in my room. It’s distracting me from my work. I can’t stop watching its trajectory, the way it moves smoothly in one direction, and then abruptly jerks to the side. It sometimes moves in circles, sometimes in straight lines, sometimes up and sometimes down. I try to understand its motion, understand why it’s flying the way it is. It’s not moving towards any food source, it doesn’t seem to have any purpose. It’s just tracing arbitrary spaghetti in a small cubic area in my room.

It amazes me that its motion is both predictable and unpredictable. I know with high certainty that it’s going to stay in this small cubic area, at least for the next few seconds. At least that’s what it has been doing for the last five minutes. But I don’t know if in the next second it’s going to keep moving in the direction it’s moving right now, or turn suddenly to the side. I wouldn’t bet either way.

At any given time, why does it choose to fly straight, or why does it choose to turn? From just looking at it with my eyes, it would be impossible for me to tell. It seems to have a mind of its own. But I know that if I was to look into its little brain, I would see a very predictable series of chemical reactions. Why does this fly move the way it does? Why instead does it not hover in place, or beeline in one direction until it hits an obstacle? Well, it’s been programmed this way. The physical events that direct its body and brain are entirely controlling its behavior.

I’ve always imagined a device that could measure every single variable of a physical system, know every law of physics that dictates its development, and put two and two together to predict the future. If you knew every bit of information that affected the flip of a coin, from the initial velocity and torque of the finger flick, to the air pressure at every atomic position around it, to the exact distribution of weight around the coin, you would be able to tell accurately whether it would land heads or tails. You would be able to take a random event, and make it deterministic.

Maybe such a device could never actually exist. There are some (quantum) theories that say the world is random all the way down to its basic building blocks. This seems to imply that you would never in fact be able to measure all the variables with 100% confidence.

But even in the light of this potential unpredictability, the world is still quite predictable. If I place my hand on the table my computer is resting on, I can predict, with exceedingly high confidence, that my hand will not pass through it. Of course, there’s an infinitesimally small chance that the atoms in the table and the atoms in my hand will align just right to make this happen. But I’ve learned enough of the world to bet that this won’t happen, and this kind of knowledge is what allows me to navigate the world effectively.

Even if the world is inherently unpredictable at its core, having the ability to know the most likely future is still enormously useful. After all, this is exactly why we have brains. Intelligence is the ability to build a model of the world, and confidently predict it. At every moment, our brains allow us to anticipate what’s coming next, so that we can affect it.

Our brains are less-powerful versions of the device mentioned above. Since they cannot model every one of the infinite variables that direct physical activity, they concede to model the most obvious ones. For our purposes, that’s good enough. Knowing that touching a fire will burn us without having to actually do it every time keeps us alive. Of course, with our capacity for intelligence, we do so much more than just staying alive. But with all the wonderful things humans can do, we’re limited by our brains. I can’t model the motion of this fly beyond a certain point, in the same way that I can’t predict the weather as well as I would like.

But what if we could build better brains? More powerful versions of these future-predicting devices? What if we could build something that could tell me, with very high confidence, exactly when this fly was going to turn in the next few seconds? Then we could turn this more powerful predictor against things like weather and stocks and the effects of public policy, things that even the smartest of us don’t have the capacity to really understand. Imagine how powerful such a thing would be. I would go so far as to say it that eventually, it could be the most powerful thing in the world.

That’s why I’m so fascinated by Artificial Intelligence. The big dream of AI is to understand our own minds, and build even more capable ones.

Today, this little fly eludes me. It reminds me how little we understand everything, with all our capacity for intelligence. But our greatest ability has always been to turn the light of our intelligence on itself. If we get this right, we can build things that allow us to finally exceed our biological grasp, and fully understand and change our world.

Nothing in life is to be feared, it is only to be understood. Now is the time to understand more, so that we may fear less. Marie Curie