Decentralized Socrates (Part 15)
Rhetorical verification would involve a large amount of text analysis by some AIs. It is possible to create a blockchain and decentralized organization to handle this analysis work.
The work and resources could be paid for, and the validation of the rhetoric verification algorithm could be done by other entities than those in charge of the analysis. The technology that forms the basis of this area is already in operation within several blockchain projects as mechanisms for decentralizing computing resources and various types of validation.
Ideally, we hope to create a system in which everyone understands that the contents of information, including its meaning, cannot be biased under the control of any particular entity.
Starting from the literal identity without touching on the semantic content, the assurance of validity becomes more difficult in the following order: no manipulation of impressions by rhetorical manipulation, no mistaken meanings or facts, and validity of arguments and inferences. “The concept of (decentralized) assurance of validity” is still an unexplored area, including whether and to what level it can be effective. It is significant, nevertheless, that only by stepping into this realm can a “degree of objectivity” emerge in the context of subjective evaluations and intentions often filled with endless arguments and relativism.
The existing guarantees are at the pure data level, such as transaction ledgers and document data. In recent years, however, technological innovations in natural language processing have made it possible for AI to extract the semantic content of individual documents to some extent. Under limited circumstances, many tasks have already passed the Turing test. This progress implies that decentralized content verification should eventually become possible in the semantic layer as well.
The rhetorical verification is the most simple place to start, as it can only attempt to detect bias in the “narrative” without going into the content.
Any method of rhetorical verification may lead to unjustified results. Suppose an AI trained on data containing many racist statements becomes a chatbot repeating such words. It is not incorrect as the behavior of the program. However, it is not justified. The algorithm + data pair is not appropriate. Therefore, the ideal system consists of various rhetorical verification AIs, and the source code for execution, data, and execution results should be transparent and re-verifiable.
Such a mechanism does not currently exist. It is also challenging to create it immediately due to the lack of computational resource maintenance, knowledge of semantic storage, communication formats and protocols, appropriate UI, and economic motivation. Nevertheless, the most complex part in principle, namely, to make the computer understand the semantic content of each document somewhat, is rapidly being solved.
As mentioned in the previous article, the actions of a particular platform to label something “by its own judgment or algorithm” can be interpreted as large-scale position talk if transparency is not ensured. Supporters of former US President Trump have argued for this reason against the attempts of companies (such as Facebook and Twitter) to ban him from SNS.
The operation of existing companies relies on the decision-making from only a few management teams in reality. Proving to a skeptic that they do not have any bias is impossible. A slightly clever fake news author can generate as many new phony stories as they want, fueled by the fact that their documents are flagged.
Removing this potentiality of the policy shift by a few decision-makers (which may be arbitrary or benign to others) from existing organizations is very hard. This difficulty is because the centralized organization is originally a method to enlarge the will of a few decision-makers.
Therefore, in principle, the rhetorical verification can only be done in a decentralized organization. The centralized organizations cannot overcome the “unreliability of the referee” described in the previous article. The verification requires “ to be known as unchangeable by the will of any particular party.”
This is the essence of “what only a decentralized organization can do,” already latent when the first blockchain, Bitcoin, was proposed as “a mechanism to prevent someone (arbitrarily or with good intentions) from deciding monetary policy.”
In the case of the rhetorical verification, the subject of legitimacy assurance through decentralization is abstracted from “the appropriateness of the amount of currency issued” to “the absence of bias in the information.” It can be said, though, that the scope of what is “ hard to justify in theory to entrust to a particular authority” is just being discovered anew.
As we will see later, despite being manufactured, decentralized organizations can behave as a kind of new “nature.”
If the “verification of semantic content” mentioned earlier is conducted in a decentralized organization, a new domain of “meaning and intention as natural phenomena” will be available in conjunction with its validity level. For the author, this is probably the conceptual significance of a decentralized organization. In a field that has been polarized into natural phenomena and subjective opinions, an intermediate field could arise in some structured form. Of course, philosophy, law, other humanities, and social sciences have created similar intermediate domains with natural language. At that time, however, forced searches for evidence to bring them closer to natural science, words that frequently appear in analytic philosophy such as “intuitively natural,” or in legal judgments such as “socially acceptable,” made their objectivity dubious. This doubt eventually led to the “unreliability of the referee. Because there are too many cases where intuitive naturalness and social common sense have been exploited for arbitrary and conspicuous distortions.
At the time of this writing, for example, there is an experiment with a blockchain-based news site that uses timestamps to check for data tampering.
Also, although slightly different, blockchain can play the role of file data storage that cannot be censored or deleted by anyone. The news that volunteers uploaded the data of the Ringo Daily, which was forced to cease publication and delete data by Hong Kong’s State Security Maintenance Law, to a blockchain to store the files permanently on a self-organized basis is fresh in our memories.
Besides “preservation of value independent from national currency policies” and “speculation and investment that does not require centralized authority,” these events will inform the public of unique ways of utilizing decentralized organizations. It exists for the “preservation of information whose authenticity cannot be guaranteed in principle by centralized institutions.”
The essence of the philosophy behind decentralized organizations is the guarantee of truthfulness. Concrete examples such as the preservation of value and the permanent storage of files are by-products of this zero trust. Looking at it in this manner, the idea of decentralization still contains a vast amount of change, just as the “preservation of value” through Bitcoin has attracted quite a large scale of investment and speculation solely by its potential.
If decentralized organizations are simple “organizations whose every decision depends on voting,” they may end up as “cumbersome organizations whose decision-making is only slow.” For this reason, we should ask ourselves, “Why do we need decentralized organizations? or “What can only decentralized organizations do?” in addition to the technical feasibility of its realization. There is still enough ground to ask these questions.
Unfortunately, the concept of the decentralized organization still appears to be exploring areas that are unique to it. We hope that a time will come when the above questions lose their worth.
Furthermore, the rhetorical verification comes with a nice bonus.
The following article will discuss this topic.