Many companies, even ours, have a tendency to throw around the term machine learning without making sure everyone understands what we actually mean. So here is our attempt to offer a plainspoken explanation and, hopefully, next time that mysterious phrase comes up, you can finally be the one who says, “Well, actually…”
Simply, machine learning looks for patterns and then validates whether or not those patterns are meaningful. It other words it looks at a huge pile of things and then gives each thing a label. Then it sifts through all those labels and makes sure each of them sit in their own pile. Finally it begins to see how these smaller piles may be connected or affect each other.
And the more things it labels and organizes the more connections it can make.
But there is nothing here that humans can’t do—it’s simply a question of scale. Any algorithm is no more than an order of operations, a way to get from point A to point B. When you go to the grocery store to pick up cake ingredients, that is similar to computer collecting data. An algorithm is a precise, highly complex recipe that tells you how to make a very specific kind of cake. This is where it becomes an issue of time and scale.
There is nothing an algorithm can do that can’t be done by humans. But depending on what needs to be done, getting the same result may require thousands of people working for thousands of hours. That’s unfeasible and why only with machine learning have so many of these results finally been reached.
But what is the difference between artificial intelligence and machine learning? They are technically interchangeable, but for a better understanding here’s an analogy:
There is a car without a steering wheel that you have to get to a mechanic one mile down the road. You could attempt to push it. It might go straight but it won’t be fast and you may get too tired to even make it to the end. Artificial intelligence, broadly, is like you getting in a car and jamming a stick against the gas pedal. You won’t get tired and you’ll go much faster but with no steering wheel you may end up in a ditch. Machine learning, in this analogy, is a tow truck. You would call a tow truck to quickly, efficiently and safely pick up the car and take it where it needs to go, because that is what tow trucks have been created to do.
How it works is a computer looks at all those small piles of things and makes a lot of little guesses (millions of guesses) about the connections between them until it feels it has come up with the most informed guess on if/how they are actually connected. It’s trial and error on a massive scale, and just like your brain’s neurons are constantly firing to make small and large decisions, the computer is doing the same.
But just because machine learning mimics the brain doesn’t mean it is a replacement for the brain. A calculator saves you from having to do long multiplication problems, and machine learning saves you from having to sift through all of this data and try to find the same kind of information. But unlike a calculator, you don’t end up with a definitive answer. Instead it is doing all the laborious research for you to offer the most expansive and in-depth look at your data and its connections. Machine learning significantly reduces your risk, helping you make the most informed decisions possible.
After all, humans built the pyramids. But it took centuries. And you don’t have that kind of time.