Here’s a classic example. Let’s say you want to train a machine to recognize pictures of cats. So you feed it two data points: It might conclude that a cat is Or what about this one? a furry thing with pointy ears, almond-shaped eyes, whiskers, and paws. But what happens = cat when it comes across this image? JStone / Shutterstock.com catwalker / Shutterstock.com = cat This is a cat, but without the telltale pointy ears or paws. Using only two images to “teach” a machine how to identify a cat probably wouldn’t equip that machine to accurately classify Garfield as such. However, feeding it billions of different images of cats — in nearly every color, size, and position possible — would make it much more likely to correctly classify an image like this: AI for CRM: Everything You Need to Know 9