How can we better understand machine learning? What is it exactly? If You have ever done any computer programming, then tyou’ll have a better chance of understanding this article.

First, a bit of background. What is an algorithm? Algorithms are at the heart of computing and are a step-by-step description of how to solve a problem. An algorithm is a recipe for solving problems. You can express an algorithm in many different forms. This first form is simply writing it out in ordinary prose or natural language. You could create a flow chart. You can use pseudocode. You can also use an actual programming language like Python or C#. A recipe for a dinner dish is an type of algorithm.

Let’s Compare

Let’s compare traditional programming to machine learning. With traditional computing we have information inputs (text, numbers etc.) that are fed into a human-written computer algorithm to produce an output. The algorithm is written by people and then coded into a language to be fed into the computing device, which might be a laptop computer, a mobile phone, tablet, microwave oven and so on.

Machine leaning is different. It takes the inputs and outputs in the form of a dataset or series of datasets and then creates algorithms based on the data. The algorithms are machine designed. It finds patterns, similarities and differences in the dataset. Typically the datasets are very large.

What about the future? The way we are now training students in AI is changing. It’s getting less technical and we don’t need to use as much logic. We are now focusing on training the computer by using more English and less computer code. Even hardware is changing. We have computer chips that are not dedicated to running conventional human-written computer programs, but dedicated to running machine learning programs. For example, Google has a Tensor processing unit.