Artificial intelligence’s progress is staggering. Efforts to advance AI concepts over the past 20 years have resulted in some truly amazing innovations. Big data, medical research, and autonomous vehicles are just some of the incredible applications emerging from AI development.

To understand some of the deeper concepts, such as data mining, natural language processing, and driving software, you need to know the three basic AI concepts: machine learning, deep learning, and neural networks. While AI and machine learning may seem like interchangeable terms, AI is usually considered the broader term, with machine learning and the other two AI concepts a subset of it.

It’s likely that you’ve interacted with some form of AI in your day-to-day activities. If you use Gmail, for example, you may enjoy the automatic e-mail filtering feature. If you own a smartphone, you probably fill out a calendar with the help of Siri, Cortana, or Bixby. If you own a newer vehicle, perhaps you’ve benefited from a driver-assist feature while driving.

As helpful as these software products are, they lack the ability to learn independently. They cannot think outside their code. Machine learning is a branch of AI that aims to give machines the ability to learn a task without pre-existing code.

In the simplest terms, machines are given a large amount of trial examples for a certain task. As they go through these trials, machines learn and adapt their strategy to achieve those goals.

For example, an image-recognition machine may be given millions of pictures to analyze. After going through endless permutations, the machine acquires the ability to recognize patterns, shapes, faces, and more.

A well-known example of this AI concept is Quick, Draw!, a Google-hosted game that lets humans draw simple pictures in under 20 seconds, with the machine-learning algorithm trying to guess the drawing. More than 15 million people have contributed more than 50 million drawings to the app.

How do we get machines to learn more than just a specific task? What if we want it to be able to take what it has learned from analyzing photographs and use that knowledge to analyze different data sets? This requires computer scientists to formulate general-purpose learning algorithms that help machines learn more than just one task.

One famous example of deep learning in action is Google’s AlphaGo project written in Lua, C++, and Python code. The AlphaGo AI was able to beat professional Go players, a feat that was thought impossible given the game’s incredible complexity and reliance on focused practice and human intuition to master.

How was a program able to master a game that calls for human intuition? Practice, practice, practice — and a little help from an artificial neural network.

Deep learning is often made possible by artificial neural networks, which imitate neurons, or brain cells. Artificial neural networks were inspired by things we find in our own biology. The neural net models use math and computer science principles to mimic the processes of the human brain, allowing for more general learning.

SOURCE: https://www.roboticsbusinessreview.com/ai/3-basic-ai-concepts-explain-artificial-intelligence/