From the course: Artificial Intelligence for Business Leaders

Build machine learning platforms

From the course: Artificial Intelligence for Business Leaders

Start my 1-month free trial

Build machine learning platforms

“

- It might seem like a lot of these artificial intelligence tools are trying to accomplish very different things. Natural language processing converts the human language into something that a computer could understand. Automated decision-making makes analytical decisions based on historical trends and data. But the reality is from the perspective of an AI system, these tools in concepts are actually very similar. With both of these systems, you're looking for patterns in massive datasets. With natural language processing, the system is looking for patterns in the way that people communicate. Then it looks at a massive database of similar patterns to see if it can find a match. With AI decision making, the system is looking for patterns of previous transactions. So when a flight gets canceled, the system knows that when this happens, they reroute to a different flight. Both of these systems rely on massive datasets and sophisticated pattern matching. That's why both of these systems use very similar concepts to accomplish these tasks. The most popular way to develop these systems is by using machine learning. Machine learning is essentially a set of algorithms that helps AI systems see patterns in massive data sets. Then the machine learning algorithms typically label the data so it can be classified or clustered together. Now it may surprise you, but machine learning has actually been around for nearly 70 years. Recently, it got a big boost because search engines like Google and Bing use machine learning to classify webpages. So when you type the word cat into a search engine, it searches it's massive database to see which web pages it's labeled with the keyword, cat. These machine learning algorithms crawl the web, so they can classify billions of webpages based on keywords. These algorithms are a bit like giant labeled guns, organizing websites the same way you might label your spice cabinet. But again, even the one system classifies web pages and the other tries to make decisions or understand the language, the underlying concept is very similar. Data is being labeled based on the machine learning's classification algorithm. With natural language processing, you labeled different phrases, so the system knows how to respond. So if you say something like, I don't understand, it can label that text as a question, then connect it to the most popular response. So you can think of machine learning as a broader concept that helps many other AI tools label massive data sets. In many ways, this data labeling is a key ingredient to make these AI systems work well. But much like any other key ingredient, machine learning can be mixed and matched with other things or just used on its own. Later, we'll take a look at what might work best for your organization.

Contents