My colleagues at Dataninja asked 3 questions on machine learning. Here are my answers:
What is machine learning?
Many of the tasks accomplished by computers result from the execution of a list of explicit instructions given by a human programmer. Machine learning is a family of techniques to teach machines do things by showing them examples, without giving explicit commands.
It works approximately in this way: we give examples (data) to a “learning algorithm”, which is a program able to extract statistical patterns from data and compress them into another little program, called “model”. The model represents the world reflected by the data and is able to make predictions and reply to questions.
Why is it useful?
To automate human jobs and lessen fatigue. We often see machine learning at work in recommendation engines (e.g. Amazon’s suggestions), sentiment analysis (e.g. Tweet categorisation), robotics (e.g. the Roomba), but it is already present in many industrial and commercial applications. It is an additional layer in the vast digital ecosystem we are surrounded by, aimed at making our lives easier by means of smarter machines.
Please keep in mind that data is at the root of everything. Tech corporations are able to build and put at our service advanced artificial intelligence because we offer them a continuous flow of personal data.
Why everybody is talking about it?
The most common machine learning algorithms were already well understood by the 80s and 90s of the last century, but they need a lot of data to work properly. Only in the latest years the digitization of society led to an explosion in quantity and quality of data. The more is hard to make order, filter and exploit the data we produce, the more we need to assign the menial work to machines and leave for us the important decisions.
Machine learning has a dark side, being the substitution of human labour, privacy erosion, power centralization, and a progressive amount of responsibility held by machines. I predict in the next years we will see machine learning discussed not only in technology debates, but also in economics and politics.