Machine Learning (ML) is poised to be the most important general-purpose technology of the coming decade. In the past few years, it has been described as "more important than fire" by the CEO of Google. One of the pioneers of ML, Andrew Ng, calls it "the new electricity". ML is a field of artificial intelligence that has immediate applicability for Japanese businesses. At its core, ML is about using labeled data to train a model and obtain a prediction. Those predictions can be applied to numbers, images, or text. It is more powerful than traditional data science, in that it is not meant to produce insights from data, but instead automate predictions.
Machine learning has been taking off exponentially in the past few years, for several reasons including: the explosion of unstructured data (images, sound, text); the availability of open source ML algorithms in repositories such as GitHub; the ease of using pre-trained models or services on platforms such as Amazon Web Services (AWS), Google Cloud, and Microsoft Azure; the continued growth in low-cost cloud computing to be able to train and run these models; and finally the availability of freely accessible knowledge on the topic from its “creators” in platforms such as Coursera and Udacity. This has been a perfect confluence of conditions to create this entire field from scratch in the past few years.
Japan has an opportunity to leapfrog other nations, by placing ML in the middle of all its future applications, rather than using traditional IF-THEN rules to govern application logic. Oftentimes, there is confusion on what these technologies are or do, and this chapter aims, through clarity, to induce action. This white paper discusses many use cases that employ traditional ML and its sub-domain, deep learning.
Much has been written about AI-ML automating jobs and resulting in job losses. This requires some qualification. AI-ML's primary applicability is to specific “tasks” rather than to entire jobs; it is best used to automate discrete actions that are hard for humans but easy for computers. These can include setting the prices for thousands of products or identifying if an image of many products is showing a defect. It is harder to automate entire jobs. However, there are some jobs, for example driving trucks on a highway from point A to B, that may be fully automated in the future, and will require re-skilling; this comes with the good news that humans continue to have their own supercomputer that can learn anything and has yet to be replicated – the brain. Lastly, it is important to note that computers can provide answers but cannot yet ask questions, and as long as that is the case, humans will retain the edge in creativity.
Comments