What is Machine Learning and why it matters?
Machine Learning, a subfield of Artificial Intelligence to not be interchanged with it, is the science of feeding computers with observations and data, for them to perform a specific task without using explicit instructions, but relying on patterns and inference instead. It has been the talk of the town for the past years. Everyone in every industry is looking forward to computers acting like humans and being able to learn and improve without being explicitly programmed.
Face detection, speech recognition and translation to text, self-driving cars, handwriting to printed letters, disease diagnosis, planes without pilots, and many other examples are based on Machine Learning algorithms, to be able to accomplish tasks without the need of human’s intervention. Here‘s a Machine Learning use-case of how the U.S. Army is using Machine Learning to predict when combat vehicles need repair.
What are the benefits and the challenges of Machine Learning?
Let’s start from one point that should be highlighted: The enemy image building should stop once and for all. The technology was founded to help humans in their tasks and facilitate their lives, but still can be used by thieves, by criminals, as well as to destroy cities during wars. It has never been Technology vs Humans; it is either technology being used for what is better for Humans, or for what is worst.
Now once we have agreed on the above, let’s get back to what are the added-values, the advantages, the disadvantages and the challenges of Machine Learning, the very powerful extension of human brainpower, founded to assist humans in their tasks and not replace them:
Machine Learning has the capability of identifying patterns from large volumes of data, which may not be apparent to humans, with ease. It frees us from repetitive tasks. It continuously improves and its accuracy/efficiency improves over time. New high-tech real-world products will be available: security drones, autonomous cars, antiviruses being able to implement new filters in a response of new threads without the need for human intervention, etc. A computer running the show itself.
On the other hand, Machine Learning needs enough good quality data to bring results. It needs resources and time for training. It has a higher level of error susceptibility, less ability to error diagnosis and correction, and limitations in prediction: They cannot always provide rational reasons for a particular prediction or decision. They have no context understanding and are limited to answering questions rather than posing them. Unlike humans, computers are not good storytellers.
Therefore, Machine Learning which is about solving a specific type of problems isn’t appropriate for every business, and some challenges should be considered before looking forward at the benefits of implementing it. Ethics for ex. are one of the major challenges to be considered.
More details about Machine Learning (algorithms and examples) in the next post.
Computer Engineer • Entrepreneur • Blogger