by Alvaro Cruz May 4, 2020
Let's take a crash course in data science terminology.
Machine learning (ML), or the study of computer algorithms that improve automatically through experience, has demonstrated vast potential for applications across a variety of industries and fields of study. Considered to be a subset of artificial intelligence and related to computational statistics, the latter’s focus mainly being prediction via computers, ML algorithms build mathematical models on training or sample data to automatically make future decisions without explicit oversight or human supervision. Algorithms may inherently seem complex, but they’re simply a collection of rules or instructions for a problem-solving process or computation, generally done by a computer. Perhaps it is this interchangeability in rules and outcomes that make them so applicable across a variety of problems and situations.
Of all the things made possible by machine learning, perhaps the most tangible is self-driving cars. Organizations like Google and Uber are actively pursuing autonomous vehicle research and development, and others like Tesla are already implementing their current technology. Other examples of how machine learning is used in practice include computer vision, or classification, detection, restoration, and segmentation of images by machines. Another is natural language processing or computer understanding of human speech and text. These examples may give the impression of recent advancements via ML as science fiction made a reality, related to technology not necessarily applicable yet in everyday situations. However, most of us encounter ML regularly without knowing it.
Image credit: https://www.nasa.gov/image-feature/active-regions-on-the-sun
Personalized ads on Google and Facebook use ML to predict what content may be most impactful to the individual. Netflix was an early adapter of ML and held a competition as early as 2006, awarding a cash prize to the best recommendation algorithm based on their then publicly made data. Eleven years later, in 2017, an estimated 80% of TV shows on Netflix were discovered through their recommendation system, the one they were looking to optimize years prior. From a business perspective, it is an intelligent thing to do to not only retain viewers or customers but also introduce them to new content or product. ML is not only useful technologically but also when it comes to savvy business practices.
One of the most potentially impactful developments of ML is in fusion energy. The sun and other stars are natural fusion reactors, their stellar nucleosynthesis, or fusion of two or more atomic nuclei into a heavier nucleus and releasing energy. The idea of harnessing fusion power as renewable energy has been around for quite some time; however, until now, there’s always been more energy expense than energy output in fusion reactor prototypes, rendering them unproductive and not commercially viable. Yet given recent advances in ML, several groups from Massachusetts Institute of Technology to TAE, a fusion company in southern California, to a 35-nation project in France estimate to be only a few years away from commercialization and viable reactors! As a renewable source, it produces less radioactive nuclear waste compared to traditional fission nuclear energy, hence the continued interest and research into its eventual production in a hopeful near future.
These are just a handful of countless examples of how machine learning can be effectively used to bring to life ideas and concepts that until recently were thought to be science fiction. From savvy business practices to harnessing the power of stars, the future applications of machine learning are bright.