Machine learning increasingly affects our digital lives—from recommending music to translating foreign phrases and curating photo albums. It’s important for everyone using this technology to understand how it works, but doing so isn’t always easy. Machine learning is defined by its ability to “learn” from data without being explicitly programmed, and ML-driven products don’t typically let users peek behind the curtain to see why the system is making the decisions it does. Machine learning can help find your cat photos, but it won’t really tell you how it does so.Last October, Google's Creative Lab released Teachable Machine, a free online experiment that lets you play with machine learning in your web browser, using your webcam as an input to train a machine learning model of your very own—no programming experience required. The team—a collaborative effort by Creative Lab and PAIR team members, plus friends from Støj and Use All Five—wanted people to get a feel for how machine learning actually “learns,” and to make the process itself more understandable.OK, but what are inputs? Or models? An input is just some piece of data—a picture, video, song, soundbite, or article—that you feed into a machine learning model in order to teach it. A model is a small computer program that learns from the information it’s given in order to evaluate other information. For example, a feature that recognizes faces in the photos stored on your phone, probably uses a model that was trained on inputs of many photos containing faces.Let’s say you train a model by showing it a bunch of pictures of oranges (the inputs).