“Explain like I’m five” or ELI5, is an idea that started on Twitter (#ELI5) and spawned the Explain Like I’m Five subreddit. ELI5 takes a complex idea or concept and goes through the mental exercise of simplifying it in a way that a young child can understand.
In this BENlabs take on ELI5, we’ll hit on some common — and not so common — terms relating to marketing, influence, and AI and do our best to distill them into simple terms without losing too much of the nuance.
Define Predictive AI in ELI5
Predictive AI learns from what has already happened to make a good guess about the future. The more it sees, the more data it gets, the more it learns, and the better its guesses get until it’s almost always right.
If I roll a ball on a flat surface, you can predict it will go in a straight line but not where it will stop. That’s because you don’t have enough data. If you watch closely as I roll the same ball three times on the same surface, at the same speed, you can “predict” where it will stop on the fourth roll.
If we make the surface wavy, what will happen isn’t as easy to see. You might need to watch 10 times before you can make a good guess.
Now, what if we make the surface flat again, but we add an obstacle like a ramp? The ball will bounce and become even less predictable. You might need to watch 10 times but on the eleventh, you can make a good guess. If you watched closely and attentively, you could probably tell where the ball would stop on the one hundredth roll.
This is basically what happens with predictive AI; it watches what happens closely, it learns with machine learning, and it makes better and better guesses until it’s not really making guesses at all, it’s able to tell us what will happen.
Because AI doesn’t get bored or tired, and because it doesn’t forget things, we can add in more obstacles. Eventually, the predictive AI will be able to make an accurate prediction, even if we alter the circumstances.