What’s the difference between AI and machine learning? Well, artificial intelligence (AI) refers to computer systems designed to perform tasks typically attributed to human beings. The term was coined in 1956 by John McCarthy at Dartmouth College. Examples include programs that translate text into other languages or play games on sites like stellarspins.
Machine learning uses algorithms to teach computers how to complete specific types of tasks without being explicitly programmed. Machines can learn from data collected through sensors or user input. They can also analyze data in previously unseen ways.
A Look at AI in Detail
When it comes to artificial intelligence, there are three main approaches: symbolic, statistical, and neural networks. Each has its own set of benefits and drawbacks.
However, let’s take a look at an example using each approach.
The first generation of AI used mathematical models called “statistical decision trees.” These were developed by IBM researchers in 1950. It relied on simpler forms of logic and worked better with large quantities of data. Today’s AIs use deep learning – a branch of machine learning — to help them reach their goals. We’ll talk about deep learning later.
In short, AI involves machines that mimic some aspect of human thought processes, such as classification, problem-solving, prediction, or inference.
Computer scientists use statistics to create algorithms. These are based on rules of thumb — things they think might work well for a given situation. For instance, if you wanted to model raindrops falling on your backyard pond, what factors would be important?
The number of drops; the size of drop; time it falls; wind speed; temperature; etc.? You could apply those same principles when building an algorithm for classifying plants. In this case, the number of leaves, leaf shape, texture, and color would all influence the results of the analysis.
Today’s advanced AIs are called “neurons” because they resemble biological neurons. Neural network technology helps machines recognize patterns and relationships. Some applications include natural language processing, image recognition, robotics, autonomous vehicles, and speech.
Neural networks don’t rely on preprogrammed instructions as do classical AIs. Instead, they adapt over time and learn from experience. This means they evolve by becoming increasingly good at performing a task.