What Machine Learning Really Is
For humanity, trying to accomplish mundane and unnecessary tasks shouldn’t be a priority at all. The view that some outdated tools are vital for our jobs according to our high school and college education, downplays the fact that there’s a huge window of opportunity in today’s technological advances that can effectively provide us with the necessary time to devote our attention to what matters the most at the end: thinking, creativity, and innovation.
One of the best inventions and state-of-the-art techniques to solve problems in many different fields, especially in businesses, is Machine Learning.
Machine Learning basically consists on a set of different algorithms (i.e., different detailed instructions that a computer can execute) that can be used to solve complex problems that we humans couldn’t solve due to cost and time constraints.
Different popular and well-known Machine Learning algorithms that are used by businesses include, for example, gradient descent, the normal equation, and different tools such as linear regression (be it simple or multiple depending on the number of variables), logistic regression, and much more.
Mainly, these incredible tools help to save a lot of time that would otherwise be lost if different calculations were done manually (by hand). Of course, Machine Learning is not only about computing calculations. It is not only software.
Machine Learning, as the concept explains it, is about teaching a computer (a machine) how to learn by itself. The idea is simple. For example: given a training set with correct and incorrect answers for some input (for example, different emails and if those are classified as spam or not), can help a computer determine and predict in the future, with a random training set, and without the correct answers, if these new emails can be classified as spam or not. This is basically called Supervised Learning (which can be divided into classification and regression)
Within Supervised Learning, we can have different tools for predicting the value of different values. This is widely used in fields like Economics, where Statistics plays a central role on this profession’s tools (e.g., we may want to predict how much the national income increases or decreases in percentage given a change of 1% in unemployment). With Machine Learning, we can implement linear and logistic regression smartly with other concepts such as vectorization (to aid our computers to compute data with less involved costs or lost time), and regularization (if overfitting is present in our model for the given data). Therefore, computers will be able to predict data on their own, saving us the necessity of losing more precious time doing ordinary calculations.
We also have Unsupervised Learning, in which computers are basically free to classify data or group information based on common characteristics and labels it creates itself (data isn’t labeled from the beginning).
Other applications of Machine Learning can extend to Neural Networks and Deep Learning, where the richest applications can be found, and most of the advances regarding Artificial Intelligence (AI) are taking place nowadays.
We hope that, in the near future, we are able to find answers to some of humanity’s deepest concerns regarding life and meaning through advances in Machine Learning and AI. And it seems we’re closer than ever to find out the mysteries of our existence.