If there’s one thing that has become as clear as day in recent years, it’s that machines will play a far more important role in the future than they already do now thanks to the recent advancements in artificial intelligence (AI).
Some of the most commonly mentioned use cases for AI include image recognition and classification, efficient processing of patient data, predictive maintenance, object detection and classification, prevention against cybersecurity threats and fraud detection, sensor data analysis, voice and image search, and smart cars, just to give a few examples.
Recently, AI made headlines when an AI-generated painting, called 'Portrait of Edmond Belamy,”sold for $432,000, becoming the first artwork made entirely by AI to go up for sale at a major art auction. But how could a machine learn how to create a work of art? It was all thanks to a field of artificial intelligence known as machine learning.
The goal of machine learning is to make machines that can learn without being explicitly programmed. “At its most basic, [machine learning] is the practice of using algorithms to parse data, learn from it, and then make a determination or prediction about something in the world,” defines this field of artificial intelligence Nvidia.
In other words, machine learning is aboutbuilding computer systems that automatically improve with experience and definingthe fundamental laws that govern all learning processes.
Machine learning is different from classical approaches to AI, which revolved around step-by-step rules written by programmers for machines to follow. Thanks to modern machine learning algorithms, AI-powered machines are mimicking the human brain and achieving things that were thought impossible just a decade ago, as demonstrated by the algorithm behind the recently sold painting.
Types of Machine Learning
Machine learning is a huge field, and there are many types of machine learning, but they can be broadly classified into two main categories: supervised and unsupervised learning.
Supervised machine learning takes known sets of input data and uses these sets of data to train a model that can makepredictions about the future based on evidence in the presence of uncertainty.Typical applications for supervised learning include medical imaging, speech recognition, electricity load forecasting, algorithmic trading,and credit scoring
Because supervised machine learning algorithms have to be first shown the right answer from a set of sample data labeled by humans, they’re sometimes compared to small children who are taught to recognize objects using repetition. A parent may show a child dozens of different images of cars before the child can reliably recognize a car.
Of course, children eventually reach a stage where they can learn how to identify new objects without receiving any help from their parents, or, in the case of machine learning, humans tasked with labeling large sets of data.
“The goal of automating machine learning is to develop techniques for computers to solve new machine-learning problems automatically, without the need for human machine-learning experts to intervene on every new problem,” says Jeff, a senior fellow with Google’s Brain Team.
The type of machine learning that is closest to achieving this lofty objective is unsupervised learning. More specifically generative adversarial networks (GANs), a class of algorithms used in unsupervised machine-learning applications.
Instead of learning from data provided by humans, GANs canfindhidden patterns or intrinsic structures in data without knowing the right answers beforehand. Their applications include gene sequence analysis,credit card fraud detection, market research, and object recognition.
Science fiction and even some people in the tech industrylike to depict artificial intelligenceas a potentially gravethreat to humanity. The reality is, at least for the time being, not nearly as alarming, but it’s certainly just as exciting. Machines have learned how to do many things in the last decade, and we can thank machine learning for it.