Learn about match patterns.
- Deep learning and machine learning have been supercharged with the availability of massive amounts of data. If you wanted to create an AI program to identify pictures with cats, then you'll typically have access to millions of these images online. You could feed your network and help it learn. It's the same with other types of data. You can easily get digital art, music, and documents. Just a few decades ago, it was extremely difficult to get a few thousand digital images or music.
Now it's pretty trivial to get access to all types of data. This new treasure trove of data has really given machine learning a huge advantage over symbolic reasoning. It's now much easier to have a neural network chew on terabytes of data than it is to have experts hard code reasoning and answers. Remember that machine learning feeds on data as a way to learn new things. The more data you feed into the network, the easy it will be for the machine to find subtle patterns.
In some ways, it's similar to how technology evolved on the World World Web. When the web first started out in the early 1990s, there were just a few sites that created directories of the entire web. One of the most successful was Yahoo. In many ways, Yahoo was like an expert system. Experts had to hard code each new site to fit into their large directory. In the early days of the World Wide Web, you would submit your website to Yahoo and then work with them to put your site in the best category.
It was a labor-intensive process because human beings had to create the categories and then manually add new sites. Sometimes Yahoo had to break down categories into subcategories, again a human task prone to error and time consuming. This process works fine when you have thousands or even hundreds of thousands of websites. It starts to break down pretty quickly when there are billions of new webpages. Human beings just couldn't keep up with that influx of new information.
So, in the mid 1990s, Yahoo started working with a smaller company called Google that would use machines to find its own categories and connections. When people clicked on a link, Google would increase the weight of the link. The Google algorithm also put more weight on websites that were heavily linked. That way, Google could create stronger connections between popular destinations. This is the same way that you should see the connections between neuron strengthening in an artificial neural network.
Each time a visitor clicked on a link that correlated with their search, they would make the Google network just a little bit more accurate. It's no coincidence that Google is one of the companies that's most enthusiastic about AI. In many ways, their whole business has been built up using machines to interpret massive amounts of data. Rosenblatt's Perceptrons could only look through a couple of grainy images. Now you have processors that are at least a million times faster looking through millions more images.
Plus, the new deep learning architecture allows machines to find patterns in the data that just a few decades ago would have been nearly imperceptible. You have many more layers of players in your marching band. These deep learning artificial neural networks look at so much data and create so many new connections that it's not even clear how these programs discover their patterns. It's like a black box where swirls of computation and data are mixed together and agree on what it means to be a cat.
No human knows how the network arrived at their decision. Maybe it's the whiskers. Maybe it was the ears. It could be something about all cats that we humans are unable to see. In a sense, the deep learning network creates its own model for what it means to be a cat, one that humans can only copy or read but not necessarily understand or interpret. So, if you're starting your own program in AI, you have to be comfortable with the fact that your network might be sensing things that humans are unable to perceive.
Artificial intelligence is not the same as human intelligence. And even though we might reach the same conclusions, we're definitely not going through the same process. This might be fine if you're looking for images of cats, but it becomes far more of an issue if your AI program is designed to decide who gets a loan to buy a house.
This course will introduce you to some of the key concepts behind artificial intelligence, including the differences between "strong" and "weak" AI. You'll see how AI has created questions around what it means to be intelligent and how much trust we should put in machines. Instructor Doug Rose explains the different approaches to AI, including machine learning and deep learning, and the practical uses for new AI-enhanced technologies. Plus, learn how to integrate AI with other technology, such as big data, and avoid some common pitfalls associated with programming AI.
- The history of AI
- Machine learning
- Technical approaches to AI
- AI in robotics
- Integrating AI with big data
- Avoiding pitfalls