Why You Should Care About Artificial Intelligence
Unless you've been living under a rock for the last 30 years, you've probably heard of "artificial intelligence". These days, it's quite the buzz word. But what exactly is artificial intelligence, and why should you care?
Broadly speaking, "artificial intelligence" refers to intelligence exhibited by machines (similar to your personal computer, but much more powerful), in contrast to the "natural intelligence" of humans and other animals.
These days, artificial intelligence is all around us.
Have you ever talked to Siri on your iPhone? That's powered by artificial intelligence.
Do you ever use Google Maps for directions? That's powered by artificial intelligence.
Have you ever deposited a check using a mobile banking app? Watched a recommended video on YouTube? Tracked your activity with a FitBit? Seen Microsoft Word auto-correct your typo? Been contacted by your bank about potential fraudulent activity on a credit card? Used voice-to-text to send a text message?
All powered by artificial intelligence.
From healthcare to social media and everywhere in between, artificial intelligence is shaping our lives. What do all of these applications have in common? Each simulates some aspect of the human mind, such as understanding spoken words or recognizing faces. And behind each of these applications is one or more algorithms, complex sets of instructions that help make sense of the incoming information and provide meaningful results.
Yet each application uses its own unique combination of algorithms and techniques. We won't get into algorithms just yet (they're pretty complex!), but let's talk about some common techniques used in artificial intelligence.
Data Mining involves finding patterns in very large amounts of data. For example, a clothing retailer might want to improve their customer service by looking at historical data about customer satisfaction. They can use data mining to look through company records of sales and returns; customer reviews; mentions of the company on social media; etc. While it would take humans a very long time to find patterns in such a large amount of data, computer processors are powerful and can find patterns that would be hard for a human to detect much more quickly than a human ever could.
Machine Learning (ML) enables machines to learn how to perform tasks by processing a large amount of data. The machine is first "trained" with some input and the desired output. A human teaches the machine what kinds of features to look for. Then, with enough training, it can "learn" how to respond to new inputs that were not part of the training.
For example, if you are building an application that recommends musicians based on a user's listening habits, you might start by training the application with some specific recommendations, like this: someone who listens to Garth Brooks might also like George Strait; someone who listens to Deana Carter might also like Trisha Yearwood; etc. You would also teach the system about the features that make this a good recommendation: genre (country music); lead vocals (alto); decade of popularity (1990s); etc. With enough training, the machine "learns" to identify these features and can start making recommendations of its own.
Deep Learning is a particular type of machine learning that utilizes artificial neural networks to simulate the human mind. You can think of artificial neural networks as a system of complicated layers that take some input and, through many steps, transform it into an output. At each stage of the process, each neuron in the network gives a confidence score, which together determine the confidence in the output.
Say you're designing an application that identifies trees based on pictures of their leaves. You'll begin, as with machine learning, by providing training examples of images and the corresponding tree. In this case, though, there is no need to teach the system about relevant features; it learns the features as it goes! As each image passes through the layers of the neural network, the image is processed according to features that the machine identifies, such as leaf color, texture, shape, and so on. After processing, the application provides its best guess as to the tree's identity, then learns by getting feedback about whether or not its answer was correct.
These techniques are used for many different kinds of applications, such as:
Natural Language Processing - recognizing written or spoken language
Robotics - building robots that can perform tasks with little or no human involvement
Computer vision - recognizing images or videos
...and many, many more.
As you can see, artificial intelligence is not just some crazy idea in a science fiction novel; it's actually in use, today, all around us! It's important that we understand how and why artificial intelligence is being used, so that we can be informed consumers. AI is great, but it's not magic. So if someone wants to sell you a toothbrush that can do your taxes, it helps to know the technology well enough to be skeptical. Understanding how AI works can also help us get the most out of our devices. Siri isn't great at answering every question, but if you know how she works, you can learn to ask the right questions to get the answers you're looking for.
My goal is to make us all better informed about artificial intelligence. In the coming weeks, I will be providing more detail about the techniques and applications mentioned in this article, as well as others. I hope you'll join me!