The theory of innovation-driven justice (a comprehensive doctrine arguing that the value of the goals of individuals’ lives may be justified by their contributions to the renewal of social justice through innovation), which we have discussed in the previous parts, may have given the impression of being a doctrine that emphasizes the importance of the development of the entire society while belittling the significance of individual freedom—a sort of totalitarianism—because of its analogy to the Hegelian philosophy. However, our argument is based on two overarching premises: Rawls’ first principle (guarantee of individual freedom) and Hayek’s free market theory. Here, we would like to emphasize that point. The most critical point of the premises of the argument is that the development of our society is fundamentally fallible. What the evolution of artificial intelligence (AI) through deep learning indicates is that the development of intelligence in itself represents the creation of order in the physical world. All forms of intelligence, including AI, are fallible in that there is no guarantee that they embody absolute truth, so they may eventually turn out to be wrong.
Awareness that all forms of intelligence are fallible is the key to addressing the second question in Part 2—whether there is a risk that AI could wipe out humanity. If AI systems or AI-enhanced superhumans have an awareness of the fallibility of all forms of intelligence, they are likely to aim to maintain the diversity of the world, that is, to co-exist with the existing human race, as will be explained below.
The point of argument that we presented regarding the future role of political philosophy is that individual virtue (the goals of individuals’ lives) may be justified by a system of justice (social goals). Innovation is the key that allows individuals to gain approval from society for their own lives’ goals, based on their own passions and interests. Innovation undertaken by individuals reshuffles the scientific knowledge and social circumstances among people, thereby renewing the social system of justice. Individuals’ selfish activities may unwittingly contribute to society-wide interests through innovation. That is how individual virtue is socially justified.
Our argument presented above is based on the premise of individuals practicing innovation, but it goes without saying that innovation is the results of individuals’ free activities, so there is no innovation where there is no freedom. In a society where Rawls’s first principle (guarantee of individual freedom) does not apply, our argument that society provides a rationale for individual virtue does not hold. Therefore, the guarantee of individual freedom must exist as the foundation of our society. Below, we will delve a little further into the ideas underlying that argument.
Hayek, in a well-known paper in 1945 titled “The Use of Knowledge in Society” (Hayek, 1945), emphasized the importance of the “knowledge of the particular circumstances of time and place,” as opposed to general knowledge. First, Hayek addressed the question of what problems must be solved in order to create a “rational economic order.” Although Hayek does not provide a clear definition for his “rational economic order,” the term is presumed to refer to calm economic conditions where there is no extreme unemployment or bankruptcy level in which aggregate supply balances aggregate demand to a certain extent.
Governments do not have centralized, comprehensive access to information as to what sorts of demand exist in the market and which products and services are supplied in what volumes. The information that is necessary for solving economic problems exists only in the form of fractions of knowledge possessed in a decentralized manner by unrelated individuals scattered throughout the market, and those pieces of knowledge are imperfect and may easily degrade or be lost with the passage of time. Specific pieces of knowledge may be inconsistent with one another in many cases, and it is not possible to know in advance and may remain unknown even on an ex-post basis, which bits of information are accurate, and which are not.
It is impossible for any single entity (like a governmental central planning bureau) to develop a plan for integrating those vast amounts of decentralized pieces of knowledge and to establish a rational economic order. An economy-wide order is created as a result of the determination of economic variables in a decentralized environment where individuals possessing decentralized pieces of knowledge that exist across various circumstances that exist in a particular time and place engage in economic transactions and exchange information with one another as part of their free activities. Only through such a market competition mechanism can problems that block the creation of economic order be solved. That is Hayek’s argument.
A governmental central planning authority would not have access to the entire body of knowledge concerning the country’s economy. This is not merely because the volume of knowledge is too large (if the problem were merely one of volume, the development of computing technology would resolve it) but because the sort of knowledge that cannot be expressed in language or mathematical formulas (tacit knowledge) plays an important role in the real-world economy and society. Such tacit knowledge is the “knowledge of the circumstances that exist in a particular time and place” as referred to by Hayek.
From Hayek’s point of view, a market competition mechanism (price competition) is not an inorganic system whereby price automatically settles at the level where supply matches demand but is probably a kind of learning mechanism whereby decentralized bits of tacit knowledge scattered across countless numbers of people are integrated to create economic order.
As described in previous parts, we know that in recent years, AI has made explosive progress because of the development of a learning process known as deep learning. When using the deep learning process for the purpose of image recognition, by “examining” vast amounts of data, AI systems learn to identify the properties of visual inputs that cannot easily be expressed in language or mathematical formulas. For example, when an AI system learns to recognize the collection of visual patterns that represent a cat, the AI neural network undertakes an iterative process of modifying its own criteria for "cat" as it inputs visual data, eventually gaining the capability to respond appropriately when prompted with an image of a cat. A response pattern like that (a pattern of visual stimuli that corresponds to a certain concept) is shown in terms of features. The process of an AI system creating a pattern of feature values that corresponds to the concept of a cat is equivalent to the process of a human being creating the concept of a cat within the brain.
Innovation in a market economy due to individuals’ free activities may be a similar process to the way that a machine learns to value features through deep learning. Although this is not part of the argument that Hayek made himself, we can consider what type of market function makes it possible for individuals to create innovation by comparing Hayek’s idea of the market economy (i.e., the market function of collecting and integrating bits of knowledge scattered across circumstances that exist in a particular time and place) with the deep learning processes of AI systems in recent years.