A.I. Machine Learning: What is it?

Study by the Royal Academy: only 9% know what machine learning is  

So, what is machine learning, anyway? Here is a quick history and the basics:

The man credited with developing early machine learning and coining the term - in 1959, was Arthur Samuel, an electrical engineer, and scientist from Kansas, who worked for Bell Labs and IBM before becoming a professor at Standford in the sixties.

He described machine learning as "a computer science that gives computers the ability to learn without being explicitly programmed."    

He developed the basics of machine learning by writing a program that could play checkers. Source

Basic explanation of machine learning

At its most basic level, machine learning is software systems set up by programmers that sift through information inputs via three main logic gates, often referred to as neural networks. The programmer defines the desired outcome with an output model.

The learning occurs when data is run through that system over and over again until the desired outcome is achieved at an acceptable percentage. The higher volumes of data pumped in and the more diverse the variety of information, the quicker and more sophisticated the learning becomes. 

Simple example: training to recognize single digit numbers handwritten

A simple example of this could be to teach a computer to recognize a handwritten single digit number in a black and white image. Let's say the number is 8. The images appear to the computer as lighted pixels that take the actual shape of single digit numbers. 

The computer needs to know what the image of an 8 looks like. That image becomes the output or primary model that all other single digits can't be in order for the computer to accomplish its goal.

Source of infographics: 

Again, lighted pixels in a black and white image glow to represent the shapes of the numbers. (Before that the computer has to be trained to be able to distinguish between glowing or lighted pixels and unlighted or non-glowing ones).

The programmer then shows the computer black and white images with glowing pixels that represent the shapes of all single digit numbers including 8. So, it sees a 1,2,3,4,5,6,7,8,9, and 0 inputted through its neural network system. This is done over and over again.   

It compares the images of all single digit numbers to the output image of an 8 and disregards the images that don't match. The images are fed into the system over and over until it can distinguish between all other images of single digit numbers and the image of an 8 with a high rate of accuracy. From that point forward that particular software system can always recognize a handwritten image of the number 8. The same can be applied to millions of different models that can increase machine learning's capabilities across many sectors of information processing. Source