**Part I: What is Machine Learning? Combining Measurements and Math to Make Predictions**

The labels “machine learning” and “artificial intelligence” can be used interchangeably to describe computer systems that make decisions so complicated that until recently only humans could make them. With the right information, machine learning can do things like…

• look at a loan application, and recommend whether a bank should lend the money

• look at movies you’ve watched, and recommend new movies you might enjoy

• look at photos of human cells, and recommend a cancer diagnosis

Machine learning can be applied to just about anything that can be counted/measured, including numbers, words, and pixels in digital photos.

What makes it different is that, with machine learning computers don’t need humans to write out incredibly detailed instructions about how to identify bad loans, good movies, cancer cells, etc. Instead, computers are given **examples** (or goals) and **math**, and come up with their own instructions.

**The examples** used for machine learning contain “countable” things that measurements can be pulled from, like: dollar amounts in loan applications; the words in movie titles; and the colors and shapes in photos of human cells.

**The math** reveals patterns in the measurements. More specifically, over a series of programming steps computers can generate complex formulas (called “models”) tailored to identify and match useful patterns in the measurements. These formulas can then be used to make predictions. For example, machine learning can enable a computer to predict whether or not one of your cells is cancerous by matching the patterns in photos of your cells to photos of cancer cells it examined previously.

It’s possible for machine learning to find patterns that humans can’t see, create formulas too complex for humans to interpret, and make predictions that are as accurate or more accurate than humans can make. How this works will be more intuitive after the following thought experiment.

First, think about flipping a coin. You expect to see heads as often as you see tails. If you flip 10 times and count heads and tails, you expect around 5 heads and 5 tails, but probably not 10 tails or 10 heads. You already have a feel for this. Computers don’t. They need a formula to understand this.

Now, one more little step: Watch this one-minute video to take coin tossing to the next level and get an instant gut-level impression of what you’ll no doubt recognize as “the bell curve” (aka “a normal distribution”):

In the video, a bucketload of little beads are dropped into the top of a box with a clear front. The beads pass through a triangular grid and percolate down. At each new level they drop either left or right—just like a coin flip. As you would expect, each bead tends to drop left as often as right, so, at the bottom most of the beads are piled up towards the center. But some are outliers because they dropped more frequently right than left, or vice versa.

After seeing this just once your brain automatically gets what’s going to happen the next time beads go into the box—we humans are fast learners. Unlike a computer, you don’t need a formula to predict what will happen next time. But notice that the pattern of the (countable) beads also echoes the wavy line graphed across the bottom of the box—a bell curve—that is the formula for the bead drop pattern.

Here’s where machine learning comes in. Imagine a computer that’s set up to automatically count—and calculate the ratios of—the beads that land in each slot at the bottom of the box. It now has a formula to predict the proportion of beads that will fall in each slot the next time the box is used. This is the essence of machine learning. It’s only a slight oversimplification to say that whenever measurements about some kind of event can be collected, a computer can apply math to identify patterns and calculate a formula for predicting how a new event of the same type will turn out.

The beauty of machine learning is that it can be applied in situations vastly more complicated than coin flips or bead drops. It can discover formulas that we humans couldn’t possibly come up with on our own because, unlike the bell curve in the example, many patterns in nature are too subtle and complex for our brains to pick up on. And machine learning can operate much more quickly and consistently than humans can.

As its use expands rapidly, more and more companies are employing data scientists to help develop and improve custom machine learning solutions. Meanwhile, basic machine learning options are already built into software people use every day, like spreadsheets and CRM systems.

Next time: What’s the difference between machine learning and artificial intelligence?