In this blog post, we’ll examine what the advancement of artificial intelligence means for humans through the films ‘The Bicentennial Man’ and ‘I, Robot’, and reflect on the limitations and risks of today’s rapidly advancing AI, as well as its potential for coexistence with humans.
A household robot is delivered to a family. Based on three principles—“A robot may not injure a human being or, through inaction, allow a human being to come to harm; a robot must obey the orders given to it by human beings except where such orders would conflict with the First Law; and a robot must protect its own existence”—it begins living alongside the family. The robot’s name is Andrew. He is designed to make life more convenient by cooking, cleaning, and caring for children. However, as time passes, the robot begins to try to understand human life, and the humans living with him also come to recognize his uniqueness. This robot teaches himself by reading books and strives to become more like a human. Living with one family across three generations, he experiences the process of losing his loved ones one by one and gradually comes to desire recognition not as a machine, but as a being in his own right. Eventually, through technological advancements, he acquires human organs, begins to feel emotions, and falls in love with a woman. In the end, he gives up eternal life and meets his death, thereby gaining recognition as a human being. This is the plot of the movie ‘The Bicentennial Man’. Watching the movie made me reflect on artificial intelligence once again. If we define artificial intelligence as technology that implements human capabilities—such as learning, reasoning, perception, and natural language understanding—through computer programs, then Andrew in the film can be viewed as a machine equipped with a form of artificial intelligence. The movie portrays the most ideal form of such artificial intelligence: Andrew contributes to the advancement of human society by developing artificial organs through his learning capabilities, falls in love, and lives harmoniously among people. However, it is worth considering whether we can be certain that real-world artificial intelligence will develop solely in such a positive direction. Today’s artificial intelligence goes beyond simply executing predefined commands; it demonstrates a wide range of capabilities, such as understanding and generating human language, analyzing images, writing code, and solving complex problems. In particular, the latest AI systems—including large language models—are achieving results that rival human capabilities in various fields, and in some narrow domains, they even outperform humans. However, this does not mean that AI thinks in the same way as humans or possesses a human-like sense of self. While demonstrating remarkable capabilities, the latest AI systems still exhibit “hallucinations”—generating false information as if it were fact—and show limitations in consistently understanding and judging situations; they also differ from humans in their ability to continuously learn and adapt on their own. The OECD also assesses that while the latest AI systems possess significant capabilities in various fields, such as language and problem-solving, they have not yet reached a level of consistent and reliable human-level performance. So what does it mean that machines can evolve through learning and make their own choices and judgments? As AI advances further, we can imagine the possibility that it might think and act like humans. In a sense, one might even expect machines to make more rational choices than humans. This is because, while humans frequently make mistakes and, depending on the situation, may let emotions override logic, artificial intelligence can make judgments based on specific criteria using vast amounts of data and take into account information that humans are prone to overlook. On the other hand, the judgments made by artificial intelligence do not always yield the best results for humans. This is because the results calculated by AI do not necessarily align with human values, emotions, or social contexts. Ultimately, depending on the goals assigned to AI, the data it learns from, and the criteria by which it is designed and used to make judgments, this technology can be either a benefit or a detriment to humanity. Humans have achieved tremendous progress while living on Earth, but at the same time, we have lost a great deal. Environmental pollution is worsening, frequent wars have claimed countless lives, and crimes of all kinds are occurring. In an extreme scenario, one could imagine a situation where, after analyzing vast amounts of information about human behavior and society, a machine concludes that human existence or behavior hinders the achievement of a specific goal. However, this possibility cannot be equated directly with a “scenario where AI dominates humans,” as depicted in movies. The current risks posed by artificial intelligence go beyond the mere notion of machines rebelling against humans. Issues such as AI generating incorrect information, making biased judgments, infringing on personal privacy, exerting excessive influence on human decision-making, or impacting real-world society in unforeseen ways are also being treated as significant risks. In fact, the U.S. National Institute of Standards and Technology (NIST) is establishing a risk management framework to enhance the reliability and safety of AI, and has recently emphasized the importance of post-deployment monitoring—continuously observing AI systems deployed in real-world environments to identify unexpected outcomes. Since AI is a human-made system, errors can occur, and it may produce unexpected results depending on the training data, design methodology, and operating environment. Therefore, as AI advances, the crucial question is not simply “Will machines dominate humans?” but rather “How will humans control AI and use it responsibly?” Furthermore, as AI becomes increasingly autonomous and interconnected systems evolve to the point where they can perform certain tasks without human intervention, the question of how far humans should intervene and under what circumstances they should halt the system’s operation will also become critical.
This brings to mind another film: “I, Robot,” directed by Alex Proyas. Like “Bicentennial Man,” it involves a robot equipped with artificial intelligence, but this film depicts the extreme scenario mentioned earlier. It depicts a central AI controlling robots—originally developed to serve humans—that concludes humans are unnecessary and attempts to dominate the human world. We cannot definitively assert that such cinematic scenarios will not occur once AI advances to that level. However, current AI is far from being an entity with a single, unified consciousness and purpose that controls all of human society, as depicted by the central AI in the film. The AI risks discussed today are instead focused on specific problems that could arise as various AI systems become increasingly involved in human work and social decision-making. Since AI systems can produce unexpected outputs or create new risks even after being deployed in real-world environments, there is a growing consensus that testing during the development phase alone is insufficient; performance and risks must be continuously monitored during actual use. After watching the movie ‘Bicentennial Man’, we can consider another scenario. What would happen if such an AI robot were to develop emotions like a human? What if it laughed when happy, shed tears when sad, and these emotions were reflected in its value judgments? It is difficult to conclude with certainty that this would lead to positive outcomes, as depicted in the film. Humans experience a wide range of emotions. We feel not only joy and sadness but also jealousy and anger. In the human world, human emotions and desires often influence the occurrence of crimes, ideological conflicts, and wars. These are things that can happen simply because we are human—such as feeling jealous of someone more accomplished than ourselves, feeling anger toward someone who has hurt our pride, or refusing to accept those who hold different views. Viewed from this perspective, the scenario in which a machine goes beyond merely mimicking human emotions to actually possessing an internal state corresponding to those emotions raises yet another question. However, it has not been confirmed that current artificial intelligence actually experiences human-like emotions. While AI can analyze human emotional expressions and generate language and behavior that match them, expressing emotions and actually experiencing them are two distinct matters. Therefore, just because AI speaks and displays empathy like a human does not mean we should immediately conclude that it possesses human-like consciousness or emotions. Rather, an important issue moving forward is not merely whether AI possesses emotions, but also the possibility that humans might interpret AI’s responses as genuine emotions or intentions. As the relationship between AI and humans grows closer, this distinction may become even more critical, because humans tend to easily ascribe anthropomorphic meanings to entities with which they converse. Ultimately, we need to distinguish between the fact that AI can appear and converse like a human and the fact that it is actually a human-like entity.
As time passes and technology advances, people have begun to seek greater convenience. From the development of the telephone and the computer, through the internet and cell phones, to smartphones in recent years, people are enthusiastic about new technologies. Yet, even today—when life is far more convenient than in the past—people still pour out complaints about how inconvenient things are. There seems to be no end in sight to today’s technological advancements. Artificial intelligence is undoubtedly a groundbreaking technology that will make human life more convenient. However, as mentioned earlier, it is a sensitive technology with many factors to consider—one that could actually cause greater confusion if we focus solely on convenience. The development of artificial intelligence is no longer confined to the stage of simply imagining future possibilities. Generative AI is already being utilized in various fields, such as writing, translation, programming, information retrieval, and image and video production, and the scope of AI’s integration into actual work processes and social systems continues to expand. Consequently, discussions are becoming increasingly important not only regarding how far to advance AI but also about which areas to restrict its use, what criteria to use to verify safety, and who will be held accountable when problems arise. In fact, the European Union is phasing in the AI Act, and the implementation and enforcement of key provisions are already underway. In particular, transparency regulations—such as requiring notifications that users are interacting with AI or labeling content generated or modified by AI—have begun to take effect. This trend does not signify an attempt to unconditionally halt the development of AI; rather, it demonstrates that the deeper a technology penetrates human life, the more it requires appropriate accountability and oversight. Therefore, it is worth pausing to consider whether the continued growth and development of this technology is unconditionally beneficial. If AI evolves in a way that complements human capabilities and makes life more convenient, it can be of great help to humanity. Conversely, if humans become overly reliant on AI’s judgments or accept AI-generated results without sufficient verification, new problems may arise. Perhaps the belief that all machines can live in harmony with humans—as Andrew did in the movie—is overly optimistic. That said, assuming—as in ‘I, Robot’—that AI will ultimately dominate humanity is also an excessive form of pessimism. The crucial issue is not merely whether AI will replace humans, but rather the principles and standards under which humans should develop and use AI. While determining the appropriate limits for AI development remains an extremely difficult problem, I believe it is time to start considering where those limits should lie.
