In this blog post, we will examine how machine learning, big data, and artificial intelligence technologies will transform future society, exploring their application areas and societal impact.
What is Machine Learning?
The emergence of AlphaGo in 2016 served as a pivotal moment, widely informing the entire nation that machine learning—specifically the deep learning method—which had previously been heard only in academic circles, had already surpassed human capabilities. Governments and companies, having realized the power of machine learning, reportedly began developing a Korean version of AlphaGo, albeit belatedly. As we enter a hyper-connected society where people and objects are all closely connected through the development of the internet, it is necessary to understand what machine learning is, its concepts, types, and prospects, as it has emerged as the new mainstream in the IT field.
Concept and Characteristics of Machine Learning
To understand machine learning, one must first grasp the concept of big data. Big data generally refers to the storage of information possessing distinct scales in terms of velocity, volume, and variety that are incomparable to conventional data. Machine learning is the most practically utilized method among the tools for understanding and analyzing this big data to derive new meaning.
Machine learning differs from data mining in terms of the nature of the data it handles. While data mining typically deals with structured data in a predetermined format, machine learning has the advantage of being applicable to data in more flexible formats. For example, when dealing with an animal, data mining might handle pre-classified items like species, gender, and age. Machine learning, however, is not constrained by these specific items like species, gender, or age. It can autonomously add new items as needed, such as size or weight.
Furthermore, machine learning is similar to artificial intelligence (AI) in that it solves problems autonomously. However, it advances beyond traditional AI methods by being able to react and adapt to new data. Furthermore, while traditional AI emphasized deterministic algorithms based on conditional statements and logical expressions, machine learning has advanced by being most similar to statistics and effectively explaining stochastic aspects.
Application Areas of Machine Learning
Machine learning is driving innovation across diverse fields. For instance, in healthcare, it analyzes pathological data to aid in the early detection and prediction of diseases like cancer. In finance, it is used for stock price forecasting, fraud detection, and personalized financial product recommendations. In manufacturing, machine learning enhances the efficiency of quality control and maintenance tasks. In transportation, such as autonomous vehicles, it helps analyze road conditions and find optimal routes in real-time. Additionally, in customer service, chatbots are evolving to answer customer questions more accurately and quickly through machine learning.
Types of Machine Learning
Machine learning is broadly categorized into supervised learning, unsupervised learning, and reinforcement learning.
First, supervised learning assumes labeled data, meaning the correct answers are known. For example, identifying common features across various car photos to classify new data as cars is the result of learning from data where the “car” classification has a known correct answer. This approach can be developed to improve the accuracy of classifying data against this known correct answer.
Second, unsupervised learning, unlike supervised learning, is based on probabilistic models without predefined correct answers. It only indicates which data points are proximate and related; interpreting what concept the clusters of proximate data represent is left to human interpretation. For example, applying unsupervised learning to the phenomenon where private conversations on SNS decrease whenever a social event occurs in the media shows that the association between Argentina and Korea is very low, while a strong association appears with Uruguay. While the unsupervised learning results themselves do not output the reason for this, researchers can interpret it as stemming from geographical proximity or rivalry.
Finally, reinforcement learning draws inspiration from the network structure of neurons interconnected within the human brain. It enables efficient problem-solving through a process where each algorithm shares, evaluates, and improves its results. Though it has quite an ancient origin, its practical application only saw rapid growth after 2006 and remains the most talked-about topic today. Notably, deep learning, a leading representative of reinforcement learning, succeeded in distinguishing which photos correspond to cats through continuous learning of image data, even though it is not supervised learning where object concepts are predefined. This distinction goes beyond merely recognizing faces composed of eyes, nose, and mouth, but also identifies features down to the pixel level—the lowest unit invisible to the human eye—thus surpassing human limitations.
The Social Impact of Machine Learning
The advancement of machine learning extends far beyond mere technological achievements, exerting profound influence across society. In the labor market, increased automation and efficiency simultaneously lead to the decline of traditional jobs and the creation of new occupational forms. For instance, while automation on production lines reduces simple repetitive tasks, new professions like data analysts and AI specialists are emerging. Furthermore, as machine learning technology is applied to political and economic spheres, data-driven decision-making becomes increasingly crucial. While this can enhance the transparency and efficiency of policy decisions, there is also a risk that misuse of data or biased algorithms could exacerbate social inequality.
The Future of Machine Learning
As machine learning algorithms continue to advance, humanity is poised to achieve significant success in solving future challenges through coexistence and collaboration with artificial intelligence. While historically focused on engineering problems, machine learning will increasingly be applied to diverse and complex evolving social issues in the future. This process will emphasize the importance of interaction between humans and machine learning. Furthermore, research is underway not only in implementing machine learning using computers but also in developing computers physically suited for machine learning. Since deep learning ultimately draws inspiration from the connections between neurons that compose the human brain, it is highly likely to evolve in a similar form.
