You may have heard that machine learning is an indispensable element of AI. Or you might say, “Deep learning is a type of machine learning.” It is already used in various fields such as predicting user needs and recommending products, and forecasting product demand in stores. With further technological advances, the need for products equipped with machine learning technology will continue to grow.
In this column, I’d like to briefly explain the definition of machine learning and how it differs from AI, which is often confused.
What is machine learning?
Machine learning is the core technology of today’s AI (artificial intelligence), and deep learning is also a type of machine learning. Sometimes referred to as Machine Learning.
To put it simply, the outline of machine learning is a technology that “makes the machine itself find regularity and judgment criteria in the data (= learning), and makes a hypothesis based on it and executes it.” The regularity and judgment criteria here are: “The number of passengers is higher at night than in the morning on weekdays.” “People who bought machine learning books in the past also bought books on AI (artificial intelligence) afterwards. It’s easy to do. ”
“Learning” and “inference” in machine learning
“Learning” is a series of work from investigating the characteristics of data to creating a “learning model” that serves as a judgment criterion after the investigation, and then deciding the combination of multiple data. For example, when searching for images of “smartphone” and “computer” from a huge amount of images, features such as the size of “smartphone” and “computer”, the presence or absence of a logo mark, and the position of the camera are derived. From among them, we will quantify the characteristics that can be judged as “this is a smartphone” and “this is a personal computer”, and determine the optimum data combination to distinguish the images of “smartphone” and “computer”.
As a result of the training, an “inference model” is constructed from the actual data. Using it, we will select images with the characteristics of “smartphones” or “computers” from a huge amount of images. This work is “reasoning”. In this way, in machine learning, we make judgments and predictions according to the purpose from the data. A huge amount of data is required to discover the characteristics of a “smartphone” or “computer” by machine learning. Insufficient amount of data can interfere with the inference process.
What machine learning can do?
If you want to apply machine learning technology to your actual work, consider it by dividing it into three uses: “identification,” “prediction,” and “execution.” For example, in response to picking up the purchasing group that you want to be the main target from the customer list (identification) and considering what kind of product you are likely to purchase, including the content of promotion activities (prediction). Machine learning technology can be applied. In addition, applications have already begun for standing in stores to serve customers and driving trains (execution).
In addition, the work of finding unsolicited emails from received emails, classifying emails by email address when receiving emails, and the face recognition system installed in smartphones are actually applications of machine learning.
The use of artificial intelligence in business can be said to utilize new regularities and judgment criteria created from data for prediction and identification. However, it is up to humans to ultimately decide where and how to use machine learning in their business. In addition, even if machine learning itself is introduced in the field of business, human support is still needed in terms of accuracy.
What is the difference between machine learning and AI?
When “machine learning” is taken up in various media such as the Internet, “AI (artificial intelligence)” is a word that is used in the same way. What is the difference between machine learning and AI (artificial intelligence)?
AI (artificial intelligence) can be broadly classified into two types: ” Artificial General Intelligence (AGI)” and “Specialized Artificial Intelligence (NarrowAI)”. To put it simply, general-purpose artificial intelligence attempts to reproduce “thinking and acting like humans.” For example, customer service robots and nursing care robots, which have recently become a hot topic in the news, can be said to be applicable.
On the other hand, specialized artificial intelligence is about letting machines do some of the work that humans think and perform. By learning from the accumulated past data, it is possible to solve problems and work independently in a specific field. Most of the AI (artificial intelligence) currently in use is included in specialized artificial intelligence.
And machine learning is one of the categories of specialized artificial intelligence because it is a technology that analyzes data collected from a specific event, finds out features and rules, and makes judgments and predictions. In addition, machine learning has been studied as a field of AI (artificial intelligence) and has a history of developing by machine learning alone. Furthermore, since the development of machine learning has also raised the level of AI (artificial intelligence) technology, it seems that it is often regarded as “machine learning = AI (artificial intelligence)”.