In this blog post, we’ll examine the technologies and key elements required to implement strong artificial intelligence that can make its own judgments, act independently, and even understand emotions.
From self-driving cars, AlphaGo, and fingerprint recognition devices to the robots that think and act independently as seen in movies, all of these technologies are based on artificial intelligence (AI). As AI has rapidly advanced, it has become deeply embedded in various sectors of modern society, and its scope of application continues to expand. The various smart devices and automated systems we use today are also designed based on AI, with the ultimate goal of making human life more convenient and efficient.
However, fully autonomous robots that think and act on their own—as depicted in movies—differ significantly from current AI. This is because most of today’s AI consists of systems designed to analyze given data and perform specific tasks. So, what elements need to be added to move beyond machines that simply execute commands and create robots capable of making their own judgments and taking action? To understand this, we first need to examine the types of AI and the direction of technological development.
First, let’s examine the difference between weak AI and strong AI, which represent the most significant distinction in the stages of AI development. The fingerprint recognition devices and self-driving cars mentioned earlier fall under weak AI. Weak AI refers to artificial intelligence that lacks self-awareness and can perform only specific tasks according to the objectives and algorithms designed by humans. In other words, it is AI created to solve specific problems or tasks.
AlphaGo, which we are all familiar with, is also a prime example of weak AI. AlphaGo was designed with an algorithm specialized for the specific domain of Go; while it demonstrated abilities that surpassed even the highest human skill levels in Go, it cannot solve problems in other fields. As such, narrow AI is optimized only for specific tasks and is utilized as a tool to fulfill a single purpose.
Narrow AI can be further divided into general narrow AI and narrow AI with expanded versatility. General narrow AI performs tasks only within a specific domain. For example, systems that classify spam emails or detect cancer cells in medical images fall into this category.
In contrast, narrow AI with expanded versatility can acquire new information based on its learning capabilities and apply it to different situations. For instance, it can observe and learn how a person prepares a dish they’ve never made before and replicate it exactly, or learn and perform new tasks. Of course, current AI has not yet reached the level where it can freely learn from all situations like a human, but its versatility in performing various tasks is continuously improving thanks to advancements in large-scale language models, multimodal AI, and reinforcement learning. These developments are considered crucial steps toward achieving strong AI. At this stage, the ability to go beyond simple, repetitive tasks to learn new information and respond appropriately to situations is required, so various AI technologies—including machine learning and deep learning—play a key role.
In contrast, the AI from the “Terminator” films, the robots from “I, Robot,” and Jarvis from “Iron Man” are often cited as examples of self-aware strong AI. Strong AI refers to artificial intelligence capable of making independent judgments and taking action based on its own self-awareness, with the goal of achieving general intelligence on a level comparable to that of humans. Such AI possesses not only reason but also emotions and the ability to understand context, enabling it to make judgments and act on its own without specific commands. The primary distinction between weak AI and strong AI lies in autonomy and the ability to make independent judgments; only when these capabilities are in place can a machine act based on its own judgment.
Developing strong AI requires both a level of reason and emotion comparable to that of humans. While many people do not question the necessity of AI possessing reason, opinions vary on whether emotion is also essential. Reason is a crucial element for analyzing given data and making rational judgments, while emotion plays a vital role in understanding complex human social contexts and emotional situations beyond mere information analysis.
While current AI is rapidly advancing in its ability to analyze and make judgments based on vast amounts of data, it has not yet reached the level of experiencing actual emotions or understanding complex emotions like humans do. For example, chatbots or AI assistants can generate empathetic responses and carry on conversations that match a given emotional tone, but these are based on learned patterns rather than actual emotional feelings. Nevertheless, since human decision-making often involves the interplay of reason and emotion, there is a growing consensus that both reason and emotion are crucial elements for achieving human-like strong artificial intelligence.
The potential and necessity of developing strong artificial intelligence are currently topics of active discussion in both academia and industry. If human-level strong artificial intelligence is realized, it has the potential to bring about significant changes across society. In particular, it is expected to play a crucial role in supporting human decision-making or solving complex problems across various fields, including healthcare, education, law, scientific research, and the environment.
For example, in an aging society, AI can enhance the accuracy and efficiency of healthcare services by assisting medical professionals with diagnoses and treatment planning. It could also analyze vast amounts of data and propose optimal solutions to address climate change and environmental issues. In this way, strong AI that collaborates with humans has the potential to become a technology of immense social value.
To advance weak AI into strong AI, we must first develop its current rational judgment capabilities to a much higher level. While existing expert systems can process vast amounts of data and analyze various situations based on rules, they have limitations when it comes to understanding new concepts or thinking creatively, as humans do. To overcome these limitations, various AI technologies—including machine learning, deep learning, reinforcement learning, world models, and technologies to enhance reasoning capabilities—must advance in tandem. Through these technologies, AI will be able to learn rules on its own, adapt to new environments, and think more flexibly.
In addition, genetic algorithms—a representative method of evolutionary computation—can also play a significant role. Rather than seeking a single correct answer, genetic algorithms find the optimal solution by repeatedly evaluating and improving numerous candidate solutions. As an algorithm that mimics natural selection and the evolutionary process, it is utilized as one of the most effective methods for solving complex problems.
Endowing strong AI with emotions is also a critical challenge. Since human emotions are highly complex and manifest differently from person to person, there are limitations to simply programming AI to express specific emotions in specific situations. Current AI systems can generate emotional expressions appropriate to a given situation, but they do not actually feel or experience those emotions. Therefore, to develop AI that communicates more naturally with humans, research must continue to deepen our understanding of how humans form and express emotions, as well as their social interactions, and incorporate these insights into the systems. As this research advances, AI is expected to be capable of more natural and human-like interactions.
If human-level rationality, emotion, and autonomous judgment are all realized in this way, strong AI could become a reality in the future. However, such strong AI has not yet been actually implemented, and most AI today still falls under the category of weak AI. If AI technology continues to advance, strong AI has the potential to evolve beyond a mere tool to become a vital partner that collaborates with humans to solve various social problems. At the same time, however, challenges such as safety, ethics, legal liability, and human controllability must also be addressed and discussed.
