Machine Learning: Counting Jellybeans the Easy Way
Homer: Kids, there’s three ways to do things. The right way, the wrong way, and the Max Power way.
Bart: Isn’t that the wrong way?
Homer: Yeah, but faster!
People frequently associate machine learning with the Max Power way: just like humans, only faster and better. But its real benefit is its ability to uncover relationships between variables that are virtually impossible for humans to dream up, and then use those relationships to make predictions about the current or future state of things.
What kinds of relationships?
I have a 10-year-old son, Atticus, whose school had a jellybean counting contest, where you had to guess the number of jellybeans in a jar. Over 200 kids participated, and mine ended up winning (unfortunately the prize was the jar of jellybeans, and not a year of free tuition as I’d hoped). The jar held 1729 jellybeans, and Atticus guessed 1750.
I asked him how the heck he got so close. He said that when he went to examine the jar, as all the kids had a chance to do, he noticed there was a scale on the side of it that went up to 1700. That seemed like a pretty reasonable number to him, so he went with it, and added a bit of padding since it was filled to the top. Obviously the jar didn’t have a jellybean scale, but it turns out that by some random coincidence of nature, there’s a roughly one-to-one milliliter-to-jellybean ratio.
That’s the kind of weird, factual relationship machine learning algorithms can uncover. They can sift through hundreds or thousands of variables and reveal correlations that would otherwise take a leap of faith, uncanny instinct, or the sort of luck only 10-year-olds have. And since it’s computers doing the work, they can produce results with incredible accuracy and velocity.
Or, as Max Power would say: you don’t cuddle with machine learning – you just strap yourself in and feel the G’s!