This blog post delves deeply into the limits of control brought by technological advancement and the meaning of human existence, exploring the film 『The Terminator』, which prophesied a future where artificial intelligence surpasses human intelligence.
In the ashes of the 2029 nuclear war, the machines rise. They wage a fierce extermination campaign to wipe out humanity and send their assassin back to 1984 Los Angeles. This is the opening of James Cameron’s film, The Terminator (1984). The film’s premise is that such events occur in 2029, but is it actually possible for such robots to appear? Is it possible to create machines that are nearly indistinguishable from humans in appearance, yet physically and intellectually superior enough to dominate humanity?
I will attempt to offer my own answer to this question. I find it fascinating and enjoyable to contemplate the long-standing human curiosity about the potential of artificial intelligence and to explore creative solutions. In conceptualizing and writing this piece, I referenced books by John Searle (Mind) and Haugeland (Artificial Intelligence). These two books approach artificial intelligence from a philosophical perspective. I intend to build upon their content to examine the problem and develop a discussion from an engineering viewpoint.
Before contemplating future Terminators, let’s consider the robots already among us today. The mechanical robots humans have created have already surpassed humans in many fields. Not only do they excel in computational and memory capabilities, but robots that appear to speak human-level language and robots capable of mimicking various human expressions have also been developed. Even robots that seem to express emotions like humans are being researched. Now, let’s compare these remarkably advanced robots of 2012 to the Terminator from the movies. The Terminator demonstrates capabilities nearly on par with humans, such as independently perceiving its surroundings and actively executing missions. While there are some aspects that distinguish it from humans, its overall appearance is unmistakably that of an intelligent life form. In this respect, today’s robots and the Terminator seem quite distant. It’s true that even the most advanced robots of this era look rather meager when compared to the Terminator, which is nearly omnipotent.
Scientists who once sought to develop artificial intelligence faced a paradox known as the ‘paradox of mechanical reason’. The core question of this paradox was, “Is it structurally possible for machines to become intelligent?” This doubt arose because people at the time likely struggled to imagine machines as anything more than simple ‘tools’. In the past, even the most advanced machines operated merely as ‘a combination of simple movements’. For example, while steam locomotives and spinning machines from the Industrial Revolution transformed human quality of life, their movements themselves were extremely simple. Consequently, it was difficult to consider them intelligent. It seemed as if the words ‘mechanical’ and ‘intelligent’ were contradictory.
Centuries later, the advent of computers brought a major shift to this notion. Computers performed calculations and memory functions on behalf of humans, and these capabilities clearly appeared to be attributes of ‘intelligence’. While debate remained, people began arguing that the paradox of machine reason had been resolved based on the invention of computers. Although many philosophical debates arose over the question “Are computers truly intelligent machines?”, from an engineering perspective, the opinion that they were ‘intelligent’ prevailed. Even the latest artificial intelligence theory, connectionism, proposed methods for creating computers that think in ways similar to humans. Computers certainly seemed to be getting smarter.
But was the problem of machine intelligence solved? Personally, I believe the problem of machine intelligence remains unsolved. This is because human-level artificial intelligence entities like the Terminator have yet to appear in this world. Connectionists have proposed a direction to solve this problem. But suggesting a direction for solving a problem is different from actually solving it. For the problem to be fundamentally resolved, a complete solution must emerge, going beyond merely figuring out how to solve it. In this case, ‘artificial intelligence’ itself must emerge. In that sense, only when we can truly implement an intelligent machine in this world can we say “machines can be intelligent.”
The criteria for defining ‘artificial intelligence’ here can be ambiguous. I will define ‘artificial intelligence’ as “a machine that can recognize its surrounding environment and exhibit responses to it at a level similar to humans.” This is not about philosophical questions like “robots must possess a self,” but rather an empirical issue: whether a robot actually appears to behave appropriately in response to its environment.
So why haven’t humans yet achieved true artificial intelligence? To understand this, let’s examine the previously mentioned connectionism theory. The founders of connectionism, currently the most advanced theory for implementing AI, drew inspiration from the brain’s structure—a neural network formed by countless neurons connected via synapses. They sought to replicate this structure through computer programs to create artificial intelligence. The result of this effort is the ‘Artificial Neural Network’. An artificial neural network learns by gradually adjusting each neural network node until its output value approximates the known target output value. (The term “learns” may be debatable. However, at the very least, artificial neural networks appear to be learning.) By persistently learning in this manner, artificial neural networks can solve even highly complex problems. Yet, no matter how much repetitive learning they undergo, current artificial neural networks cannot achieve the level of artificial intelligence seen in something like the Terminator.
While several issues exist, the most fundamental problem is that the computing power we possess is utterly inadequate to replicate the human brain. Not long ago, an artificial neural network capable of recognizing cat faces became a hot topic. Despite having no other function besides recognizing cats, this neural network reportedly required 16,000 CPUs as its central processing unit. Implementing the entire human brain would require an incomparably larger number of CPUs, ultimately hitting a physical limit.
So, is creating an artificial intelligence entity like the Terminator an impossible task for humans? It doesn’t seem likely to become easy anytime soon. The physical limitations are the most painful. I believe that continuing along the path of how science and technology have developed so far will inevitably lead to hitting a wall at some point. To achieve breakthroughs like AI implementation, the approach itself must be innovative.
Let’s briefly recall moments of past invention and innovation. From the steam engine to electricity and light bulbs, to the internet’s emergence, humanity has always driven innovation through paradigm shifts that transcended existing boundaries. AI, as part of such innovation, also requires a change in approach. Rather than merely trying to overcome limitations using existing methods, it is crucial to seek new paths. One such novel approach being explored by many scientists is the biological approach. This involves seeking new solutions through the convergence of existing robotics, computer science, and biology. If a purely physical solution is difficult, the idea is to find a different way around it.
The reason biology comes to mind is the functional superiority of living organisms. There are frequent instances where living organisms surpass the limitations inherent in mechanical, artificial creations made by humans. For example, the synthesis of an amino acid chain, which takes a long two years artificially, occurs within an E. coli cell in just 0.00001 seconds. The idea is to leverage this remarkable capability of living organisms to explore the implementation of artificial intelligence. I have considered three specific examples for this.
The first method involves artificially reproducing the DNA base sequences that express the brain and utilizing them for AI programming. The second method could involve expressing the brain of a living organism and using it as the brain of a robot. Finally, artificially replicating DNA to create artificial lifeforms resembling existing organisms also appears feasible. One might question whether robots created this way “can truly be called robots in the strictest sense.” However, considering the dictionary definition of artificial intelligence, this seems less problematic.
Ethical issues may also arise. I believe this is an inevitable consequence of using living organisms as tools. Since this approach involves utilizing the ‘brain’—the organ most closely associated with mind and intelligence—it could face significant backlash. However, if these issues are well-addressed and biology is effectively harnessed, this approach might become central to the birth of artificial intelligence.
We have now considered the birth of artificial intelligence and the potential resolution of the paradox of machine reason. Today’s machine robots have made significant progress from the past, yet they cannot think and act as actively as humans. This can be seen as a problem stemming from the historical ‘paradox of machine reason,’ and solving it in a strict sense appears difficult. This is because physically implementing artificial intelligence presents major challenges. In this essay, I have proposed convergence with biology as a solution to this.
Now, let us return to the initial question posed. By 2029, will humanity have solved the paradox of machine reason and created machines possessing intellectual capabilities equal to humans? The time remaining for such an innovation, akin to the advent of artificial intelligence, is not long. Therefore, it does not appear easy. However, technological and scientific progress tends to advance rapidly once certain bottlenecks are overcome. Thus, we cannot entirely rule out this possibility.
If the advanced artificial intelligence we anticipate does emerge in 2029, biology might well be at its core. And this new approach could rewrite the history of science. Alongside technological progress, humanity will face increasingly complex and sophisticated questions. Ultimately, we must solve these challenges and open the door to yet another innovation.
