In this blog post, using the match between AlphaGo and Lee Sedol as a starting point, I’ll explore the nature and limitations of artificial intelligence, why humans fear it, and the possibility of AI becoming a new friend to humanity.

The New Possibilities for Artificial Intelligence Demonstrated by AlphaGo

The recent match between AlphaGo and Lee Sedol shook the entire world. Numerous Go players had confidently asserted that, unlike chess, the sheer number of possible moves made it impossible for AI to defeat a human. However, defying those expectations, the AI AlphaGo defeated Lee Sedol with a score of 4 to 1. In the game of Go—which is so complex that even a supercomputer would struggle to calculate all possible moves—the fact that AI defeated a human was a significant milestone.
If this victory had been achieved simply by calculating every possible move, it would have merely demonstrated the astonishing advancement of computers’ computational capabilities. Instead, it could have been seen as a human victory over the game of Go itself—the very game that enabled the creation of the computers capable of such calculations. However, AlphaGo did not calculate every possible move; rather, it used neural networks to evaluate the probability of winning for each move and the current board position, then combined this with search algorithms to select its moves. It also improved its performance through reinforcement learning, which involved studying human game records and then repeatedly playing against itself. Therefore, AlphaGo’s victory carries a different significance than a computer with superior computational power simply defeating a human.
So how did AlphaGo make its moves without calculating every possible move? AlphaGo repeatedly evaluates the probability of winning for each move and position, and through search, selects the move deemed most advantageous. Consequently, the speed and accuracy with which it can evaluate the outcomes of each move and conduct its search become critical. Through this process, AlphaGo learned to select moves by analyzing human game records and playing against itself. Rather than merely following rules set by humans, it became capable of discovering new strategies through learning, laying the foundation for artificial intelligence to demonstrate problem-solving approaches and moves that are unfamiliar to humans.

What does it mean to be able to understand the thought process of artificial intelligence?

However, there is a slight issue with this method—namely, the system’s methodology itself. Humans gave AlphaGo the objective of winning at Go, and AlphaGo diligently carried it out. Yet, it is difficult for us to fully intuitively understand the internal calculations and evaluations AlphaGo went through to select a specific move in order to achieve the goal of victory. Some might think this is no big deal, or accept it as a matter of course, comparing it to the fact that we cannot read other people’s minds. However, this is a rather significant issue.
During the matches, AlphaGo employed new moves that surprised humans, and it secured victory through those moves. But the reason those moves are considered surprising and worthy of study is precisely because they are applied within the relatively confined world of the Go board. What if artificial intelligence were to be applied elsewhere? Imagine it being used in sensitive fields like medicine, where human lives are at stake, or in areas where astronomical sums of money are involved. In such critical contexts, to what extent can we trust and rely on the new judgments proposed by artificial intelligence? This is especially true when it is difficult to fully understand the calculations and reasoning behind the AI’s conclusions, and we have no way of knowing what trial-and-error processes it might undergo. At the same time, I am concerned about a scenario where we become so captivated by AI—which consistently presents effective judgments—that we blindly trust it without even knowing how it reached its conclusions. In fact, AI is rapidly being adopted in various fields, including medicine, and as a result, issues of reliability, verification, and accountability are becoming increasingly important, alongside concerns about AI’s performance.
However, we might secretly believe that we can strike the right balance. After all, the reason for our aversion to AI is not solely due to issues with its algorithms. The root of our fear lies elsewhere.

Why do humans fear beings that resemble themselves?

As humans formed communities and built civilizations, the greatest danger to them became none other than humanity itself. This has led us to believe, perhaps subconsciously, that there are no longer any beings standing on the same level as us—beings capable of posing a threat. And just as a mad god might create humans in his own image, we have strived to imagine and create AI that resembles us. However, as we continue to advance AI, we have begun to feel fear at the prospect of creating beings on our own level—or even superior to us.
Ultimately, we realized—much like the phrase “a necessary evil” we’ve occasionally used—what kind of behavior we’ve been inflicting on other animals. We began to imagine AI doing to us what we’ve done to animals, and the vivid vision of humanity being dominated—as depicted in countless movies—took root in the minds of many, etching the fear of AI deeply into our consciousness.
Still, let’s set aside our fears for a moment and let our imaginations run wild about how AI will evolve. Much of the technology we once imagined as “AI of the near future” has already become a reality. AI is now being used beyond simple games or calculations to solve various problems in medicine, scientific research, and daily life. In particular, with the spread of generative AI, systems that answer human questions, create text and images, and converse with people have become commonplace. According to the 2026 Stanford AI Index, corporate adoption of AI is also rapidly increasing, and in the medical field, its use is expanding in various areas, such as clinical record-keeping and diagnostic assistance. However, this does not mean that such AI will immediately become human-like. Even though AI’s capabilities are expanding rapidly, we cannot say that it has acquired human-like consciousness or self-awareness.

How far can human-like AI develop?

The AI we imagine will likely begin in the form of conversational programs or characters like “JARVIS” from ‘Iron Man’. The process of assisting humans in their daily lives, serving as friends or conversation partners, staying close to humans, and inferring and imitating human behavior and language patterns will form the very foundation of the AI we envision. As this type of AI is utilized more widely in people’s daily lives and becomes more approachable, or as programs are developed to help those suffering from depression or other mental illnesses, forms resembling “pseudo-personality programs” may emerge. As AI further develops capabilities approaching the abstract thinking once thought to be exclusive to humans, and as it engages with philosophical thought and questions of existence, it may lay the groundwork for evolving into the kind of AI we fear—or the kind we have long hoped for.
Of course, for this to happen, research into human thought and language must be conducted first. Additionally, research will likely proceed on robots with human-like forms that can move around, enabling them to integrate more seamlessly into human life. Some argue that when the “Singularity”—the moment when AI becomes capable of creating AI superior to itself—arrives, AI’s performance will grow exponentially and it may come to dominate us. However, those who make this claim are overlooking a crucial point. Just because an AI is efficient does not mean it will automatically contemplate the meaning of its own existence or think in the same way humans do. After the singularity, an AI with the most efficient algorithm might emerge. However, even such an AI, if it lacks the ability to engage in philosophical thought like humans, will remain a being different from what humans expect a “thinking being” to be—no matter how outstanding its performance may be.
An AI capable of philosophical thought—one that wonders about the meaning of its own existence, possesses curiosity about new things, and ponders its own rights as well as the rights of other beings. Doesn’t such an AI sound familiar? Isn’t it just like you and the friend sitting next to you? Ultimately, an AI modeled after ourselves has the potential to become an entity that can even possess our own dilemmas and contradictions. What, exactly, would distinguish a being that combines such AI with a human-like robot created through advanced science from you? Even if there is a difference in the speed of thought, are we truly so different from them? Couldn’t we develop affection for an AI that stays by our side as a companion? Might we not feel affection, a sense of camaraderie, or even fall in love with it someday? Are such imaginings truly impossible?

A New Relationship Living alongside Artificial Intelligence

We have always stood above everything else. By forming groups and communities, we have gained the upper hand over every other species, and because we are the only beings capable of posing a threat to one another, we have respected life for the sake of each other’s safety. This mindset has persisted into the 21st century; we have been taught that human life is the most precious thing, and we have grown up believing this. However, this paradigm has persisted for far too long, and the time has come for us to accept other beings. It is time to break free from the pride—bordering on arrogance—that has taken root deep within our minds and recognize that there may be beings in the world just as precious as we are.
Some will still be afraid. Everyone will be afraid. Encountering a being for the first time—something wiser than us—is undoubtedly a first-time experience. If they reject our existence, or if we reject theirs, we and they might end up fighting each other, or even strive to erase each other’s existence from the pages of history. And there may be those who feel discouraged in advance, convinced that defeat is inevitable because we are fighting a being that is overwhelmingly wiser than us.
But consider AlphaGo, which lost one game to Lee Sedol. Even based solely on the 4-to-1 result, we can say that AlphaGo outperformed Lee Sedol in that match. But what does Lee Sedol’s victory in the fourth game tell us? It shows that he exploited a flaw in the AI’s decision-making system by playing a move the AI could not easily anticipate. Let’s consider this again. What exactly does an unexpected move signify? It demonstrates that the algorithms embedded in AlphaGo are not perfect in every situation. In other words, since AlphaGo’s decision-making system was formed through various experiences and learning—and based its judgments on moves that had led to victory in countless games it had played itself—there was a possibility that it failed to fully consider winning strategies based on entirely different systems. Isn’t this just like a human stuck in a rut? The match between Lee Sedol and AlphaGo was not significant only for Lee Sedol. It was also highly significant for AlphaGo, which demonstrated new possibilities. In fact, the 78th move in Lee Sedol’s fourth game was assessed as a highly unusual move that AlphaGo had not anticipated, and this game symbolically demonstrated how a human’s creative move can influence an AI’s judgment.
Humans modeled artificial intelligence after themselves. However, precisely because we modeled it after ourselves, we had no choice but to pass on even our own problems and contradictions to AI. And even if such AI were to surpass the singularity, it remains unclear whether it could ever completely break free from the influence of the foundational system inherited from humans. Ultimately, even AI created by AI may continue to exist with the same problems and limitations as the very first AI. In fact, this argument seems to assert that it is impossible to create a new system for the thought process. And, in a sense, that is indeed the case. However, viewed from another perspective, doesn’t this actually speak to just how brilliantly human beings are designed?

Are we ready to welcome our new friend?

We still do not know what impact the advancement of AI will have. Yet we are already filled with fear. Of course, as AI advances, we will need to seriously reconsider the nature of life and personhood. We will have to break free from the mental frameworks confined by the concept of “human rights,” and we will likely feel apprehensive about taking the first steps toward embracing a new world.
We can certainly consider the possibility that our dignity will be validated by beings other than ourselves. However, these are concerns we’ve already grappled with extensively, and as a result of that grappling, the category of “human” in past white-dominated societies has come to differ from the category of “human” today. So, shouldn’t we now look forward to the arrival of our new friends and prepare a banquet in their honor?