In this blog post, we’ll explore the possibility of AI from the movies becoming a reality, and what opportunities and threats it might present to humanity.
Iron Man, Interstellar, The Matrix, Terminator. What these movies have in common is the presence of artificial intelligence. In these films, AI is different from the machines and objects we encounter in our daily lives. For example, the computer I use to do my homework doesn’t understand my voice and doesn’t move on its own. It only performs functions when I issue commands using a keyboard or mouse. However, AI in movies understands human speech, makes its own judgments, and acts independently. “JARVIS” in Iron Man and the robot “TARS” in Interstellar serve as assistants, carrying out various commands from the protagonists. In contrast, the AI robots in The Matrix or The Terminator seek to destroy human civilization and even attempt to control all of humanity. Of course, this is impossible with current technology. While it may be merely a fantasy possible only in movies, we cannot completely dismiss the possibility that it could become a reality depending on future research and development. Could an era truly arrive where humanity’s survival is threatened by AI?
First, let’s examine what artificial intelligence is. Artificial intelligence can be broken down into “artificial” and “intelligence.” “Artificial” refers to things created by humans that do not occur naturally. On the other hand, the meaning of “intelligence” is difficult to define clearly. While the dictionary definition refers to “the ability to engage in intellectual activity,” scholars offer different definitions depending on its elements, scope, and characteristics. Generally, there is a common understanding that it refers to the ability to utilize past experiences and learn by adapting to new objects or situations. This is not significantly different from the meaning of intelligence as we commonly understand it. Therefore, artificial intelligence refers to machines or programs created by humans that possess the ability to adapt to and learn from new things. This artificial intelligence is classified into weak AI and strong AI based on the scope of its intelligence.
Weak artificial intelligence appears to possess intelligence but does not actually possess it. Examples include the software in self-driving cars, programs that perform internet searches via voice recognition, and programs that interact with users in simulation games. Most of the artificial intelligence currently in use or under development falls into this category. While they appear intelligent, they are limited to making judgments based solely on pre-inputted information. In other words, they are incapable of adapting to new situations or objects. For example, a self-driving car may appear to adapt to new situations by recognizing objects or people that suddenly appear on an unfamiliar road or driving path and stopping or detouring accordingly. However, these recognition and judgment criteria cannot exceed the scope predefined by the developer, and it is impossible to utilize experiential knowledge such as “Since there was a person there yesterday, I should avoid it today as well.” Therefore, it cannot be considered intelligent. Of course, there are programs capable of learning that can modify their judgment criteria by synthesizing events as they occur, but even these modifications do not go beyond the scope set by the developer. In this regard, all current artificial intelligence remains within the realm of weak AI.
Unlike weak AI, strong AI refers to artificial intelligence that actually possesses intelligence. It is AI capable of thinking and acting like humans, or even surpassing human thought and behavior. This means it can adapt to and learn from completely new situations or objects, and think independently. It is also capable of self-awareness and goal-setting. The core of strong AI is its ability to create something new through emergence. In other words, strong AI is an entity possessing intelligence virtually indistinguishable from that of humans—that is, AI with a “mind.” Although no programs or robots with a mind currently exist, there is much debate over whether such AI could be developed through technological advancements or might emerge by chance.
There are two main approaches to achieving strong AI. One is through programming algorithms, and the other is through a bioengineering approach. First, the method of developing AI through computer programming is currently being actively researched and is relatively easy to pursue.
However, this method alone makes it difficult to achieve strong AI. KAIST neuroscientist Professor Kim Dae-sik (2015) stated in a lecture at Ewha Womans University that “the resolution of language is lower than the resolution of thought.” In other words, human thought cannot be perfectly expressed through language, and programming this into machine language is nearly impossible. For example, if a program is inputted with the criterion “cats have fur,” the AI will inevitably recognize a dog with fur as a cat or judge a hairless cat as not being a cat. While inputting more criteria allows the system to distinguish cats, this does not amount to accurately identifying them but merely calculating the probability that they are cats. Since the abstract boundaries used to distinguish objects are difficult to express clearly even in human language, programming them is extremely difficult.
The biotechnological approach is an attempt to artificially create a human brain. For example, a research team led by Lulu Qian at Caltech (2011) created four artificial neurons using 112 strands of DNA. While this system was reported to be capable of distinguishing people through training, the human brain possesses over 15 billion neurons on the surface of the cerebral cortex alone, making it still incomparable. However, as time passes, it may become possible to create over 15 billion artificial neurons. Synapses, which represent connections between neurons, are also known to change over time, and if we fully understand this mechanism, we may one day be able to create an artificial brain. This leaves us with one final question: does the “mind” exist simply because there is a brain? If a mind exists within an artificially created brain, we can conclude that the realization of strong artificial intelligence is possible.
If an artificial brain is realized, this could be proven experimentally. While using humans as test subjects raises ethical concerns, experiments could be conducted on animals capable of tool-using emergent behavior. Similar experiments have been conducted in the past; in 1970, Robert White transplanted a monkey’s brain and spinal cord into the body of another monkey, allowing it to survive for nine days. Although the monkey eventually died due to immune rejection, it exhibited behavior similar to its previous self during its survival period. Of course, there are limitations since this was not an artificial brain, but if an artificial brain were to be implemented, similar results are likely to be achieved.
In conclusion, the realization of strong artificial intelligence depends on the development of an artificial brain. If an artificial brain is developed and possesses a mind, strong artificial intelligence will become feasible. However, whether such strong artificial intelligence will pose a threat to the survival of humanity is another matter. We do not view cows or monkeys as threats to human survival, and we know that humans, too, can be conditioned to conform to reality even in absurd situations. Therefore, even if AI with a mind were to emerge, it is highly likely that it would not escape human control.
