Artificial intelligence is a domain that shares many elements with data science but often has a more specific focus. Learn the differences between these two fields and which best meets your needs in this video.
- [Instructor] At this exact moment in history, when people think about data science, the mind turns inexorably towards artificial intelligence, often with humanoid robots lost deep in thought. But before I compare and contrast data science and AI, I want to mention a few things about the nature of categories and definitions. First, categories are constructs, and by construct I mean something that you have to infer, something that is created in the mind, doesn't have this essential existence. It's a little bit like, when is something comedy and when is something performance art, and when is something acting? There's nothing that clearly separates one from the other. These are all mental categories, and the same thing is true of any category or definition, including things like data science and AI. The second one is that categories serve functional purposes. A letter opener is anything that's used to open letters. I actually use a knife to open letters. On the other hand, I know a family that uses knives to scoop ice cream exclusively. And, the tool is whatever you use it for. It's defined by its utility. The same thing is true of categories. And then finally, the use of categories varies by needs. If you're putting books on the shelf, you can use the Library of Congress system, the Dewey Decimal system. I know people who stack them by size or by color, or turn them around and do it decoratively. Any of those is going to work because they're serving different purposes. And so, when we're trying to think about categories and defining whether a particular activity is AI or whether it's data science, all of these principles are going to apply. A good example of this is the question of whether tomatoes are fruits or vegetables. Everyone knows that tomatoes are supposed to be fruit, but everyone also knows you'd never put tomatoes in a fruit salad. Tomatoes go on veggie plates along with carrots and celery. The answer to this paradox, its fruit versus vegetable nature, is simple. The word fruit is a botanical term, and the word vegetable is a culinary term. They're not parallel or even very well coordinated systems of categorizations, which is why confusion can arise. It's a little like the joke about the bar that plays both kinds of music, country and western. The categories don't always divide logically or exclusively, and the same is true for artificial intelligence and data science. So, what exactly is artificial intelligence? Well I'm going to let you know, there are a lot of different statements about this, and none of them are taken as definitive. And some of them I find to be useful, and some of them I find to be less useful. There's a little joke that AI means anything that computers can't do. Well, obviously, computers learn to do new things, but as soon as a computer learns how to do something, people say, well that's not intelligent, that's just a machine doing stuff. And so there's a sort of moving target here to this particular definition in terms of things computers can't do. You can also think of AI in terms of tasks that normally require humans. Like placing a phone call and making an appointment. Or like returning an email, or categorizing text. Traditionally humans have done that, but when a machine is able to do that, when a program's able to do it, that's probably a good example of artificial intelligence. Probably the most basic and useful definition is that artificial intelligence refers to programs that learn from data. And so, you give them some data, they build a model, and that that model adapts over time. A few common examples of this are things like categorizing photos. Is this a photo of a horse, a car, a balloon, a person? And programs learn how to do this by first having lots and lots, and lots, and lots of photos that are labeled by the people as one thing or another, as a cat or a dog. But then the algorithm is able to start learning on its own and abstracting the elements of the photo that best represent cat or dog. It's also used for translations going from one language, like English, to another, like French. The use of artificial intelligence programs has made enormous leaps in the ability of computers to do this automatically. Another one is games, like Go here. It was a very big deal when not very long ago a computer was able to beat the world champion of Go. And, it was thought to be this intuitive game that couldn't really be explained. What's fascinating about that is the computer actually taught itself how to play go. And, we'll talk a little more about that when we talk about the derivation of rules in another video. But all three of these can be good examples of artificial intelligence, simply by the sorts of things it's able to do. And so, probably, this is the best working definition of AI. And, while it can include even simple regression models, which really don't require much in the way of computing power, it usually refers to two approaches in particular. AI is usually referring to machine learning algorithms, and in particular, deep learning neural networks. I'm going to talk about those more elsewhere, but I did want to bring up one more important distinction when talking about AI. And that's the difference between what is called strong or general AI, which is the idea that you can build a computer replica of the human brain that can solve any cognitive task. This is what happens in science fiction. You have a machine that's just like a human in a box. And that was the original goal of artificial intelligence back in the 50s, but it turned out that has been very difficult. Instead, you also have what is called weak or narrow, or specific, or focused AI. And these are algorithms that focus on one specific well defined task. Like, is this a photo of a cat or a photo of a dog? That has been where the enormous progress in AI has been over the last several years. So with all this in mind, how does artificial intelligence compare and contrast to data science? Well, it's a little bit like the fruit versus vegetable conundrum. Artificial intelligence means algorithms that learn from data. Broadly speaking, there's an enormous amount of overlap between our concept of AI and the field of machine learning. Data science on the other hand is the collection of skills and techniques for dealing with challenging data. You can see that these two are not exclusive. There's a lot of overlap between them, and AI nearly always involves the data science skillset. You basically can't do modern AI without data science. But there's an enormous amount of data science that does not involve artificial intelligence. And I'll say more about that as we go on in this course. If you want to draw a diagram, I personally think of it this way. If this is data science, here's machine learning, ML. There's a lot of overlap between those two, and then within machine learning there's a specific approach called neural networks. Those have been amazingly productive, and AI refers to this diffuse, not well defined category that mostly overlaps with neural networks and with machine learning. And, it gets at some of the ambiguities, and some of the difficulty in separating these, which is why there's no consistent definition, and why there's so much debate over what one thing is, and what the other one is. But I will say this. Artificial intelligence has been enormously influential within the field of data science recently, even though data science has many other things that it does. This course focuses specifically on data science, but you'll see just how much of this information also applies to machine learning, to neural networks, and even to the field of artificial intelligence.
- Assess the skills required for a career in data science.
- Evaluate different sources of data, including metrics and APIs.
- Explore data through graphs and statistics.
- Discover how data scientists use programming languages such as R, Python, and SQL.
- Assess the role of mathematics, such as algebra, in data science.
- Assess the role of applied statistics, such as confidence intervals, in data science.
- Assess the role of machine learning, such as artificial neural networks, in data science.
- Define the components of effective data visualization.