 Hi, I'm Charlotte, and I'm here to talk about artificial intelligence. Last time we talked about word vectors, let's talk about a bigger concept, vector representation, how a computer can recognize an object based on the features that it has or those attributes. For example, think about the game 20 questions. Perhaps you're starting with, is it an animal, yes or no? If it is an animal, is it a farm animal, yes or no? Is it a livestock animal, yes or no? We could ask all kinds of questions to learn about what type of animal we're trying to identify. Those attributes are the vectors that an AI system can use to identify unique individuals. So for example, dog breeds. We might have short or long hair. We might have weight under 50 pounds. We might have whether or not it's used for hunting or sporting. These representations are vectors that are often in fancy places like arrays or multi-dimensional spaces, but you can think of them on a spreadsheet. Could I ask a question and say yes or no? Those zeros and ones build a unique identifier for each thing we're trying to represent. So we can break down dogs into large and small. We can break them down into their hair type. We can break them down into what type of work that they do until we get a unique number for each one of them. All those zeros and ones add up, and the more that we see the same types of ones and zeros, the closer those dogs are together. Like our sporting dogs might all be larger dogs with shorter hair, or our toy dogs might all have smaller bodies. We can start to compare those features, those representations, to identify them uniquely. The closer the numbers are, the more related they are. And this is how computer systems break down something really difficult, like what is a dog, to break it down into individual features and then they can represent that digitally in order to build a representation of the world.