In this blog post, we’ll examine the evolution of artificial intelligence through Deep Blue, Watson, and AlphaGo, and consider both the positive changes AI will bring and the threats it poses, such as job displacement, safety, and ethical issues.
To what extent has artificial intelligence come to resemble human intelligence?
Artificial intelligence (AI) is one of the biggest social issues today. Artificial intelligence (AI) refers to technology that implements human capabilities—such as learning, reasoning, perception, and natural language processing—through computer systems. AlphaGo, an AI Go program developed by Google DeepMind, served as a catalyst that significantly heightened public interest in both “Go” and “artificial intelligence.” Looking back at the history of AI, notable examples include Deep Blue (a chess AI), Watson (a quiz-answering AI), and AlphaGo (a Go AI).
Deep Blue has long been regarded as a crucial testing ground for artificial intelligence because it demonstrated that computers’ computational capabilities could be leveraged in fields—such as chess and other board games—where the number of possible moves is vast but the rules and objectives of the game are clearly defined. In 1997, IBM’s Deep Blue defeated Garry Kasparov, the reigning world chess champion at the time, by a score of 3.5 to 2.5 over six games, becoming the first computer system to defeat a reigning world champion under official match conditions. Based on its ability to rapidly calculate vast numbers of possibilities, Deep Blue explored and evaluated the countless possible moves on the chessboard. This example demonstrated not so much that computers could perfectly replicate human thought, but rather that they could surpass human capabilities by leveraging massive computational power to solve problems with clear rules and objectives.
Watson was developed as a cognitive computing system capable of understanding questions posed in natural language and analyzing large-scale data to find answers. In 2011, it competed against Brad Rutter and Ken Jennings—the top champions at the time—on the American TV quiz show ‘Jeopardy!’ and emerged victorious. Rather than searching the internet in real time, Watson operated by pre-processing vast amounts of information, analyzing questions using various algorithms, and then identifying possible answers. In this process, natural language processing technology played a crucial role in enabling the computer to understand questions posed in everyday human language and find appropriate answers.
The recent Go-playing AI, AlphaGo, caused a major sensation in yet another way. Go is a game far more complex than chess, with a much greater number of possible moves; the number of possible board states reaches approximately 10 to the 170th power. AlphaGo played Go by combining Monte Carlo tree search with artificial neural networks, deep learning, and reinforcement learning. AlphaGo utilized two types of neural networks: a policy network and a value network. The policy network predicted which move was most likely to be chosen in the current situation, while the value network evaluated which side was more likely to win based on the current state of the board. Through this approach, AlphaGo was able to efficiently narrow down its search range by learning which moves human experts would choose and evaluating the current position, rather than simply calculating every possible move indiscriminately.
AlphaGo began by learning from game records of human experts and then improved its capabilities through reinforcement learning by playing games against itself. Instead of simulating every move until the end of the game and then checking the results, it was able to use the value network to assess the likelihood of winning from the current position. This approach demonstrated a significant difference compared to existing Go AI systems. While existing Go AI systems relied heavily on move reading and standard pattern databases, AlphaGo utilized not only move reading but also a policy network that learned move selection from human masters and a value network that assessed the state of the game. AlphaGo’s achievements demonstrated that, for problems where all information is publicly available and the objectives and rules are clearly defined, AI has the potential to match or even surpass human capabilities to a very high degree.
However, I believe it would be an overstatement to interpret these achievements as meaning that computers have fully acquired intuition and insight—domains unique to humans. Unlike Go, the problems humans encounter in daily life often lack clearly defined goals and rules, and in most cases, we cannot obtain all the information necessary to solve them. Therefore, there are limitations to equating the capabilities of AI—which has demonstrated outstanding performance in problems with specific conditions and clear rules—with general human intelligence or judgment in the real world.
What changes and new threats will AI bring?
Given the intense focus on AI technology, it is clear that the field has been growing rapidly. In particular, AI is no longer a technology confined to laboratories or specific industries; it is now being actively utilized in various fields such as search, translation, content creation, customer service, healthcare, finance, and manufacturing. AI technology will bring about significant changes in the future. There are high expectations that these advancements in AI will yield positive effects, including cost savings, increased productivity, reduced risks, enhanced personalized services, and the development of new business models. AI can automate tasks that humans used to perform repeatedly, analyze vast amounts of data to support human decision-making, and help quickly process complex information that was previously difficult to handle.
On the other hand, the threats posed by AI are also significant. This is because the advancement of AI goes beyond the mere emergence of convenient technology; it can also bring about various issues such as safety, ethics, employment, and socio-economic inequality. First, we must find ways to safely manage AI to prevent it from causing serious harm to humanity. This means assessing in advance the risks that new technologies may pose to humans and establishing safety measures. For example, just as scientists in the 1970s discussed the potential risks of recombinant DNA research and established guidelines for research safety, we need to establish standards for AI to ensure safety and accountability in step with the pace of technological advancement. As AI performance improves, we must consider not only the risks associated with the development process but also those that may arise during actual use.
Second, ethical oversight is needed for AI systems that pose potential risks. A particularly cautious approach is required when AI is utilized in the military sector. AI weapons include autonomous weapon systems capable of detecting, selecting, and attacking targets without direct human intervention. Although these weapons differ from nuclear weapons in their mode of operation and the nature of the risks they pose, there are concerns that their relatively low cost and high levels of automation and scalability could lower the threshold for the use of force and lead to large-scale casualties. Therefore, international discussions and ethical standards are needed regarding whether to entrust life-or-death decisions to machines and to what extent human control should be maintained.
Third, the social and economic threats posed by jobs being replaced or reduced due to the advancement of artificial intelligence are significant. While past automation primarily replaced human labor in repetitive and routine tasks, recent advancements in artificial intelligence are now affecting relatively high-level intellectual tasks such as document drafting, analysis, translation, and customer service.
In a 2024 analysis, the International Monetary Fund (IMF) projected that approximately 40% of global employment could be affected by AI, and estimated that this proportion could reach about 60% in developed countries. However, being affected by AI does not necessarily mean that the job in question will disappear. While some tasks may be replaced by AI, others may see increased productivity through the use of AI, or new roles for humans may be created. The key point is that AI has the potential to change not only the quantity of jobs but also their nature and the skills required.
Are professional and traditional jobs safe from AI?
A significant portion of banking, accounting, and administrative work can be automated through advancing algorithms and data technologies. The scope of tasks that used to require people to manually enter data, perform calculations, and review documents—and which are now handled by software and AI—is steadily expanding. Therefore, it is highly likely that the use of AI will expand not only in simple, repetitive clerical tasks but also in professional work that can be processed according to set rules.
Fourth, AI is transforming even specialized fields such as news writing and healthcare. AI is already being used to write news articles based on structured data and to analyze medical images and patient data. Of course, this does not mean that these technologies will immediately and completely replace journalists or doctors. The human role remains crucial in assessing the social context of news stories and determining the direction of reporting, as well as in making final medical judgments that comprehensively consider a patient’s condition. However, it is clear that as AI becomes capable of taking over certain tasks within these professions, professionals in these fields will find it increasingly difficult to avoid the impact of AI.
Fifth, there are concerns about an increase in low-wage, mass labor. As AI and automation technologies spread rapidly, companies will be able to handle more work with fewer employees, and in the process, the wages or job security of some workers may decline. Conversely, the productivity and value of workers equipped with the skills to utilize AI may increase, potentially widening the gap within the labor market. Alongside the growth of AI technology, interest in basic income is also increasing, and some argue that if traditional jobs decline significantly, we must rethink the very social and economic structures that govern labor and income. Therefore, the issue of job losses caused by AI cannot be resolved simply by expecting new jobs to be created; it will require a comprehensive approach that addresses education, career transitions, social safety nets, and income distribution.
What preparations should we make to live alongside artificial intelligence?
Just as smartphones have drastically transformed our lives and society over the past few decades, artificial intelligence is not a change that will occur far in the future—it is a transformation that is already underway. AI is not a new technology that will suddenly emerge in the next 10 or 20 years; it is rapidly spreading throughout current industries and daily life, and its influence is likely to grow even further in the future. Therefore, to prepare for an AI-driven society, we must not wait for the future to arrive but begin preparing step by step starting now.
One could argue that what sets humans apart from computers is precisely their imagination and creativity. Even if artificial intelligence can analyze vast amounts of data, perform complex calculations, and quickly assist humans in their work, it is humans who must decide how to use the technology and take responsibility for the results. We must deeply reflect on what we need to do for future generations who will live in even closer proximity to artificial intelligence. It is important not only to provide education that explains the principles and limitations of AI but also to cultivate the ability to use AI properly and critically evaluate its outcomes.
Many futurists, as well as prominent figures such as Bill Gates, Elon Musk, and Stephen Hawking, have emphasized the need to be wary of the potential risks of AI. One reason for this concern is the possibility that, should AI with superintelligence—capabilities surpassing those of humans—emerge, humans may not be able to sufficiently control such systems.
Of course, there are various views on the timing of the emergence of superintelligent AI and its likelihood, and we cannot treat this as a certainty of the future. However, given the rapid advancement of AI capabilities, discussions regarding human control, responsibility, and safety must not be postponed until after technological development has taken place.
While the advancement of AI technology offers benefits such as convenience and improved quality of life, concerns about threats posed by AI also exist. AI is a technology with two sides: while it offers convenience and new opportunities to humans, it can also lead to changes in the job market, social inequality, and safety and ethical issues. Ultimately, it is up to humans to decide how to develop and use AI. The key is not to unconditionally fear or, conversely, unconditionally welcome the development of AI itself, but to consider both its potential and its risks, developing the technology in a way that benefits humanity and using it responsibly.
