There are two types of people in the world. The people who think about data a lot, and the people who think about data a medium amount. When I first worked in AI data, I spent my days submerged in data. What parts of this image are roads? What category of things is this image showing? Or does this 3d mapping of dots mean a car or a pedestrian?
I've kept this mindset of thinking about data a lot ever since and for the most part, it's worked well for me.
We're surrounded by data in everyday life. At a more complicated level, we can use company's financial data to figure out how they're doing. On the street, we can use the speed at which a car is driving to judge if we can make it to the other side before they barrel into us. Personally, I can use the time in which it takes for me to hit a shuttle as a rough proxy for how good of a badminton player I am.
Yet, data is about capturing a specific moment in time. It tells us about the past and has an uncertain correlation to the future. The same AI training data that I once worked on could be entirely useless if it started raining spaghetti and meatballs forever.
Most people, however, are forward looking. Yes, even those people who spend all of their time thinking about the good ol' days. When they believe that the present or future will not hold up to the past, they're making an active prediction about what's going to happen next. Sure, they might not have a very specific event in mind, but they have a rough range of potential outcomes that, to them, is worse than what they experienced before.
All of that brings me to the idea of narratives and how fickle they are.
For example, I recently hurt my ankle. As I was writhing on the ground, the only data I had was that I had a lot of pain and I would need significantly more time than before to hit a shuttle. Yet, my mind created a whole narrative that I would have a hard time training badminton over the next few weeks, which would lead to a bad performance in the Yonex Adult Nationals,which would result in having a longer road to actually making it to the olympics, which means that my badminton career is effectively over.
Now, none of those things had to be true. All I knew in the moment was how much pain I was in, which has a relatively small bearing on my overall badminton performance. After all, the tournament was almost two months away and I didn't need to entirely lose the time I spent recuperating from the injury. And everything else after that is up for debate. Perhaps I perform out of my mind at the tournament. Or perhaps my US opponents are not as good as I thought. And then all of a sudden, my olympic dreams are back on.
We overweight data when we create our narratives because that's all that we can be sure about. Pain = bad things = more bad things. And so I had a really pessimistic view of the future.
Now that I'm feeling better, I have a slightly different narrative about the future. I'm going to train and hone my badminton skills more and I'm going to walk into the tournament as one of the few players who developed under the Chinese system, which means that perhaps the others are not used to my playstyle, which means that I get a good result, which increases the chances that I can become the top player in the US, which means that I have a good shot at the Olympics.
Because the starting data changed by a tiny bit, I now have a completely different result.
Narratives are how we make sense of the future, and often seem inevitable. After all, they're backed up by a long line of past data. But data rapidly loses its value as time progresses. What seems like a solid narrative can change drastically based on tiny changes in present data.
So data, as useful as it is, is a weaker predictor of the future than the actions we take in the present. What we do in the present weighs far more in determining how the future unfolds.
Maybe it's time to give a shot to being a person who think about data a medium amount.