In this modern era, where faith in human reason has crumbled, it is the arrival of artificial intelligence (AI) that gives us an opportunity to once again believe in the everlasting progress of reason. Only after the arrival of AI has it become possible for us to imagine, as a vision of a feasible future, a world where we coexist with an intelligence that transcends humans as beings that perceive the world.
There are various approaches to developing AI. In recent years, in many cases, the term AI has been used to refer to a multi-layered neural network software program modeled on the human brain and which has been enhanced by deep learning. Of course, there are other AI systems that operate on different design concepts. However, in this article, AI refers to a multi-layered neural network within a computer that has been enhanced by deep learning.
It has become clear that as a result of the development of deep learning, AI systems enhanced by deep learning can achieve high performance, and in recent years, research and development on AI has made explosive progress. However, AI has surpassed the human brain not across all fields, but only in limited fields, including image recognition, and the go and shogi board games. Nevertheless, some forecasts predict the arrival of the AI singularity within dozens of years, which refers to the situation where AI has surpassed the capabilities of the human brain in all intellectual activity.
In any case, it has actually been verified that it is possible to reproduce at least some functions of the human brain within a computer and advance the reproduced function of the brain beyond the level of human intelligence.
It is possible to create an infinitely complex neural network structure within a computer. Sooner or later, it will become possible to create a computer-based neural network that is more advanced and complex than that of the human brain. As a result, in fields such as the natural sciences, for example, the possibility cannot totally be ruled out that AI may create an association of ideas that is beyond the human brain's ability to understand and bring about new "sciences" that cannot be understood by humans. Until recently, our worldview has held that humankind stands above everything else in the world as beings who have the most complete understanding of the world and greater control over our reality than any other animal on the planet.
However, AI has proved that something can exist in this world that can transcend humans in terms of our power to reason (although it is not a God). It does not matter whether or not the AI systems being developed by IBM and Google will surpass human reason. More important than how the development of any specific AI system will proceed is that it was shown that the possibility of a "rational being" that surpasses human reason can actually exist in human society as a vision of a feasible future. If the scope of human reason is expanded by an AI system with superhuman intelligence, the progress of the "expanded reason" may not reach its limitations for a while. In other words, although the faith in reason as an omnipotent force has declined, it may be possible to restore the system of belief under a new progressive theory of history that maintains that the limitless expansion of reason will continue. That is also related to the question of whether the limitations of AI exist in the realm of human understanding.
Given that AI is a computer system, it should have limitations based on the laws of physics. Still, it is possible to create a computer neural network that is more advanced and complex than the human neural network. Such a computer neural network would be the same as the human neural network in terms of the basic network structure in that neural cells (elements) are linked by synapses, although there is a quantitative difference in that a larger number of neural cells would be linked together in a more complex way. The quantitative difference between the computer and human neural networks results in a qualitative difference in the perception of the world. That is in line with the argument made by Philip Anderson in his article titled "More is Different."
Let us compare human and chicken brains, for example. The chicken brain is the same as the human brain in that it is comprised of a neural network. The only difference is that in the chicken brains, a smaller number of cells are linked together in a simpler way. However, the chicken's perception of the world is qualitatively different from human perception. For example, when a chicken sees its own image reflected in a mirror, it mistakenly believes that there is another chicken there and tries to attack it. This qualitative difference—humans can recognize their reflection in the mirror as their own image but chickens cannot—stems from a quantitative difference, that is, the difference in the number of neural cells that constitute the brain, which also results in the difference in the way that the cells are linked together. Both human and chicken brains are the same at the basic level, i.e., in that both have a neural network structure.
A similar difference exists between AI and the human brain. AI, if equipped with a higher number of neural cells linked together in a more complex way, will surpass the human brain qualitatively, at least in some domains of reasoning. If AI becomes the driver behind the development of the "sciences," some aspects of the sciences will move beyond the reach of human comprehension.
That would be different from humans' inability to readily understand massive calculations performed instantly by computers. When a massive calculation is performed, humans can understand what is being done and use it in meaningful way because the task is of such a nature that it could be done by themselves in principle given the necessary time. However, new "sciences" that may be developed by AI will be qualitatively different in structure from the sciences that have been developed by humans, and as a result, the human brain will likely be unable to understand them regardless of the time spent, similarly to an animal’s inability to understand human sciences.
As explained above, the arrival of AI suggests the possibility that the sciences will evolve beyond the limitations of human reason. If an intelligence can evolve in a way that is qualitatively different from the sciences based on human reason, we can be assured that reason expanded through collaboration between AI and humans will continue to progress forever toward a true understanding of the world.
Such "expanded reason" enhanced by AI will continue to progress for a while without reaching the limits of human understanding. Thanks to AI, the progressive theory of history that Hegel conceived of in The Philosophy of History can be renewed as something that advocates the progress of "expanded reason" instead of the progress of human reason (albeit with the proviso that all intelligences are fallible; I will explain in detail in the forthcoming articles).