 Hi, I'm Charlotte. And I'm here to talk about artificial intelligence. Have you heard of word vectors? How about natural language processing? Today, we're going to talk about those ideas. Natural language processing is the way a computer understands what you say out loud in a regular voice. It's part of natural interaction. Your phone is very good at understanding and translating what you say out loud to written communication, even tricky homophones like pour, not having enough money, and pour, like pouring your juice, or telling a tale versus a mouse's tale. You can try this on your phone or in translating by speaking out loud and having a document right for you. Your phone does not understand the words that you're saying, but it does know associations between words. For example, llama. Lama is very closely associated to alpaca. It's also closely associated to farm. It's also associated with clothing or stuffed animals. It's not as associated with the word tire, for example. So llama can be related to a lot of different things that aren't just related to farms. But what we have to determine for natural language processing is how similar a word is in the text compared to other words that are found in text. And if you search the whole internet, you could make a lot of comparisons. It used to be on the SAT that you'd be asked analogy questions, like popcorn is to butter as fries are to what? And you might say ketchup or mayonnaise, but you probably wouldn't say tiger, right? There are words that are more closely or less closely related to one another. And we can rate the closeness of words using math. You've done this too, probably with a game called 20 Questions. But a computer asks about 300 questions. And we can sort these words like llama into categories. We can say it's an animal, yes or no. Yes. Is it a computer? No. Is it a cooking word? I don't think so. Let's go with no. How about farming? Yes. How about clothing? Also yes. So as we choose those yes and no values, we can say that words are more or less closely related to one another and give them a mathematical value. Then we can do math on them. Like man is to king as woman is to, hopefully we get queen, right? But we can do king minus man and get some value that's like royalty. And we could do royalty plus woman and come up with queen. That's how these word vectors work. They make associations and we can do math on them to find what words are related. You can play with this too with wordssimilarity.com. Give it a check out. You can even see words in different languages and how they're related. And then maybe you'll see how closely llama is related to alpaca. If you wanna get into the meat of this, word to veck is the place to go. That is more for people who are very technical, but you can play with word vectors as well.