Zou Chuanwei: Large-scale collaboration between machines is not a fantasy, it requires blockchain to introduce market mechanisms

This article was shared online by Dr. Zou Chuanwei at the Wanxiang Blockchain Public Welfare Hackathon on March 5, and the chain news was authorized by the author to release it. The author, Dr. Zou Chuanwei, would like to thank Mr. Chen Sheng for his amendments.

Alibaba's Dharma Institute listed " large-scale collaboration between machines as possible " in the " Top Ten Technology Trends in 2020 " as the fourth largest trend. Bodhidharma believes that the development of IoT collaborative sensing technology and 5G communication technology will achieve collaboration between multiple agents. Machines will cooperate and compete with each other to complete the target task, and the group intelligence brought by the collaboration of multiple agents Further magnify the value of intelligent systems.

Bodhidharma emphasizes the role of information sharing and unified control in the network of smart devices. I think the following cutting-edge issues need to be discussed:

  • Is this the only mechanism for large-scale collaboration between machines? Are there any other mechanisms?
  • As a distributed collaborative network, does blockchain intersect with large-scale collaboration between machines?

Large-scale cooperation mechanism of human society

To understand large-scale collaboration between machines, it is best to compare it to large-scale collaboration between people. Large-scale collaboration is one of the most important characteristics of human society . Without large-scale collaboration, human civilization cannot develop to this day.

The large-scale cooperation mechanisms of human society can be divided into two categories. The first category is centralized , with top-down command chain, bottom-up feedback mechanism and multi-level principal-agent relationship as the core features. Typical representatives are administrative agencies, enterprise organizations and military units. The second category is decentralized , represented by market mechanisms. There is no central plan or unified coordination in the market mechanism. While each participant works for his own interests, he promotes group interests through the division of labor and market exchanges.

Today I mainly discuss the market mechanism .

In the market mechanism, the cooperation between people is carried out in dimensions, and these two dimensions are inseparable.

The first is the information dimension . For example, before a bank lends money to an enterprise, it needs to assess the enterprise's credit qualification. There are two methods of assessment. The first is to analyze the company's financial statements and score the company's credit. After the rise of artificial intelligence and big data analysis, a lot of unstructured information can be analyzed. The second is to use market prices to infer the credit risk of an enterprise from the stock price, bond yield and credit default swap (CDS) spread. In various markets, participants trade assets based on the information they possess. Those who are optimistic will buy, and those who are not optimistic will sell. The price of the asset formed by the transaction integrates the information held by different people. In many cases, the market price information disclosure function cannot be replaced by artificial intelligence and big data analysis. Hayek pointed out in " The Road to Slavery ": "Price is the only way for economic decision makers to communicate with each other through tacit knowledge and decentralized knowledge, so that the problem of economic calculation can be solved."

The second is the value dimension . In the market mechanism, participants decide how to allocate resources based on market price signals . For example, consumers decide how to spend their own money to maximize their welfare according to the prices of goods and services; enterprises decide what and how much to produce based on the prices of production factors and products to maximize their profits. In modern society, market mechanisms are everywhere. We are in all kinds of markets all the time, constantly producing value, constantly exchanging prices, and maximizing the use of limited resources. Adam Smith vividly and profoundly revealed the role of the market mechanism in "The Wealth of Nations": "The food and drinks we need every day are not from the favor of butchers, brewers or bakers, but from their self-interest. Everyone who uses capital and labor, neither intends to promote the public interest, nor to what extent he promotes that interest. He is guided by an invisible hand to try to achieve a What he intended to achieve. By chasing his own interests, he often promoted social interests . "

Can the market mechanism be introduced in the machine network?

As mentioned earlier, there are two types of large-scale cooperation mechanisms in human society-centralized and decentralized. The " information sharing and unified control " that Bodhidharma said represents a centralized mechanism. The market mechanism is a representative of decentralization, operating in two dimensions of information and value, and the information disclosure function of market prices cannot be replaced by artificial intelligence and big data analysis. The question is, can market mechanisms be introduced in the machine network?

This question concerns the relationship between machines and people. Different disciplines have conducted in-depth studies on the relationship between machines and people from different angles.

In 1944, Schrödinger published " What is Life ", using thermodynamics, quantum mechanics and chemical theory to explain the nature of life. The fact that many people are familiar with the negative entropy of food intake comes from this book. This is to study people as machines. In 1955, the artificial intelligence Dartmouth Summer Research project proposal was openly stated: "Every aspect of learning, or any other characteristics of intelligence, can in principle be described so accurately that it can be simulated by a machine . Will try Find out how to make the machine use language, form abstractions and concepts, solve various problems that are now reserved for humans, and improve itself. "This is to study how to make machines think like humans. The current realization is mainly weak artificial intelligence represented by pattern recognition, and strong artificial intelligence is still early .

The introduction of a market mechanism in a machine network essentially allows machines to behave like people. In the future communication networks, Internet of Things, and smart device networks, can blockchains be used to allow machines to collaborate on a large scale like humans? I think the answer is yes. Some people may think that this is science fiction, or even a fantasy. This is not the case, there is such a prototype system that has been successfully operating for more than 10 years. I refer to the Bitcoin network .

The Bitcoin network has no shareholders, managers or regulatory agencies, and has more than 10,000 nodes in the world.It has been running for more than 10 years without failure and uninterrupted operation. Recently, the daily output of mining is 16.2 million US dollars, and the average daily transactions on the chain are 350,000 and 7.1 billion US dollars. .

Many people are used to looking at the bitcoin network from a technical point of view. In fact, the bitcoin network is more valuable from an economic point of view . There are complex market economic activities in the Bitcoin network. For example, miners provide computing power, work performance is reflected in Nonce, and block rewards are awarded for this. Block rewards are sent to miner related addresses through Coinbase transactions. But nodes that only provide storage space, network routing, or wallet services have no reward. When users initiate transactions, they need to provide higher fees to incentivize miners to prioritize their transactions, which is equivalent to using fees to bid for limited system resources in the Bitcoin blockchain.

Many excellent economists have studied the economics of the Bitcoin network from the perspectives of game theory, asset pricing and network economics. We also know that the underlying technologies such as cryptography, distributed networks, and distributed computing that Bitcoin relies on have been developed before Satoshi Nakamoto. Many researchers have tried to construct electronic cash through these technologies, but they have not achieved complete success. It was not until Satoshi Nakamoto introduced game theory design that Bitcoin became possible. This shows the power of economic mechanisms : in a decentralized environment, how to prevent the evil of mine work and ensure the security of distributed ledgers, guarantee network construction, operation and maintenance, and realize self-organization and self-governance. Problems that cannot be completely solved by technology have produced good results by joining economic mechanisms.

If the Bitcoin network is regarded as a prototype system of " machine network + market mechanism ", what are the successful experiences that can be copied and promoted?

The economic expression of machine network

Before discussing "machine network + market mechanism", it is necessary to express machine network as an economic issue. The machine networks I discuss mainly include communication networks (such as 5G) , Internet of Things (IoT) , smart device networks , and blockchain networks .

These machine networks have the following characteristics:

  • First, the structure is centralized (such as "server-terminal") and point-to-point.
  • Second, different types of nodes assume different roles, such as providing data, computing power, storage, and bandwidth, and have a cooperative relationship of division of labor.
  • Third, before the Bitcoin network, the machine network was basically a physical problem, not related to economics. Compared with human behavior, machine behavior can be more characterized by functions. Input variables are control parameters and environmental parameters, output variables are machine behavior, and function settings are determined by machine design.

A special problem in the construction, operation and maintenance of machine networks is that some machine networks belong to infrastructure, and it is difficult for infrastructure to charge users directly, which is voluntary or free. The typical representative is the communication network. This belongs to the issue of the supply of public goods in economics. If the network of machines as infrastructure is completely provided by the private sector, there will be insufficient supply. There are two solutions in reality. First, the government uses fiscal and tax revenue to invest in the construction of network infrastructure and provides it as a public product. The second is " cross-subsidization ". The network infrastructure is free, but applications running on it are charged. Depending on how to attract the participation of the private sector and give corresponding incentives, various public-private partnership models (PPP) have emerged in reality.

Can machines become economic subjects? The core of this question is whether the machine has its own utility and whether it can adjust its behavior according to economic incentives.

In the intersection of distributed artificial intelligence and economics, one of the main visions is the machine economy (Machine Economy) and device democracy (Device Democracy)-machines have their own utility, and they will maximize their utility through strategic optimization and become so-called intelligence Agent . But before the realization of this vision, it can be considered that the machine has no own utility, and the incentive machine is actually the controller behind the incentive machine. Of course, I can also consider the issues I discuss today as the primary version of machine economy and equipment democracy.

Human behavior is very complex. Affected by various factors, there are many different modeling methods. The mainstream paradigm of modern economics is based on the hypothesis of the economic man, which essentially uses machines to approximate human decisions and behaviors, which returns to the relationship between machines and people discussed earlier. In this paradigm, everyone has a set of values, which determine his preferences, such as whether he likes oranges or apples. Preference is used to quantitatively describe the utility function. The utility function should reflect the characteristics of diminishing marginal utility and risk preference. The goal of the economic man is to maximize personal utility under various constraints. For example, the problem faced by consumers is to choose a consumption mix under budget constraints to maximize utility. The problem faced by an enterprise is that given a revenue goal, minimize costs, or given a cost goal, maximize revenue. In short, whether it is a consumer or business problem, it can be described by an optimization problem, and there are two major elements of an optimization problem-the objective function and constraints .

It should be noted that the economic man hypothesis is an approximation of human behavior in an ideal situation. In reality, behavioral economics research shows that it is difficult for individual decision-making to eliminate irrationality, and behavior often deviates from the assumption of economic man.

The Economics of "Machine Network + Blockchain"

"Machine network + market mechanism" is inseparable from blockchain. As discussed above, the market mechanism operates on the two dimensions of information and value. Blockchain combines the attributes of information internet and value internet , and can play a role in two dimensions of market mechanism.

As an information Internet, the blockchain uses public accounts to record the flow of commodities, medicines, food, and funds, allowing upstream and downstream, and different links to verify each other, penetrate information "islands", and make the entire process manageable, but does not involve Asset or risk transfer.

As a value Internet, blockchain involves the transfer of assets and risks. However, the blockchain itself does not create value. The value comes from real-world assets and is linked to the Token in the blockchain through economic mechanisms. The blockchain functions as an infrastructure, which is reflected in the settlement of transactions , the automation of settlement , and the intelligence.

The core of "Machine Network + Market Mechanism" is the economics of "Machine Network + Blockchain" . In terms of methodology, it is necessary to penetrate the machine network to the controller behind and use the " two-step method ".

The first step: economic activities take place on the machine network. From the perspective of production, consumption, market transactions, and division of labor, we must refine the economic activities and participants in the machine network.

Step 2: Discuss how these economic activities are supported by blockchain, including the following aspects:

  • The blockchain records economic activity. This aspect uses the blockchain as a function of the information Internet. The difficulty lies in how to make the information outside the chain fidelity "on the chain" and record the economic activities and value flow outside the blockchain in a highly reliable manner. For example, blockchain + supply chain management requires safe and efficient sensing equipment to write off-chain information to the blockchain credibly. There is no perfect solution to this problem, which is one of the core obstacles of the current blockchain application, and it is not entirely a technical problem, nor can it be separated from related institutional arrangements.
  • Blockchain provides payment and incentive tools for economic activities. In this aspect, blockchain is used as the function of the value Internet. There are two core economic issues: one is the payment tool between machines, and the other is the economic model design of the machine network. The focus will be discussed next.
  • The state change of the Token / smart contract in the blockchain triggers the machine behavior. This aspect involves the transmission of state within the blockchain to the real world.

"Machine Network + Blockchain" has three important economic issues.

The first is how the machine becomes an economic subject that can be evaluated , because the premise of economic incentives is to accurately measure the contribution of the machine. Accurate measurement of contributions includes the following three aspects:
  • One is that each machine must have a unique identity and cannot be forged or modified (there is only one "me") ;
  • Second, the behavior of the machine is traceable and non-repudiation ("I" has a record of what I have done) ;
  • The third is the authenticity of the machine's identity and behavior, which can be self-certified through algorithms, without the need for manual or institutional participation in verification ("I" is my proof) .

Only the machines in the machine network with the above characteristics, their actions and contributions can be measured, and can be regarded as an economic subject that can be evaluated, or called "machine economic man".

For this, four aspects of work need to be done:

  • One is to introduce a unique ID to the machine through modules, chips and other technologies, that is, the machine's digital identity (public and private keys) , which is implemented at the bottom of the hardware and cannot be tampered with. This is similar to personal characteristics (such as face / fingerprint, etc.) that cannot be modified and corresponds to the identity of each person.
  • Second, the machine network introduces a new addressing mechanism , using the hash of the machine's public key as the addressing element, instead of the IP address, to achieve the uniqueness of the address and identification. For example, the distributed hash table (Distributed Hash Table, DHT) algorithm calculates its neighboring neighbors based on the value of the hash address. The neighboring nodes of the neighboring nodes are constantly searching to find the target. This is similar to a personal characteristic. Give you a name and use that name to find you (addressing) .
  • The third is self-certification . Any activity of the machine will be signed with the identity of the machine. Through the public and private keys and encryption mechanisms, self-certification of the machine's identity and behavior can be achieved without the need for manual and centralized institutions. This is similar to each of your behaviors, you press the fingerprint and draw to confirm, to ensure that what you do is undeniable, and at the same time, you can also confirm that you are you through the fingerprint, without having to find someone to prove (self-certification) .
  • The fourth is interaction and accounting . The machine network no longer takes featureless information packets as the main body. Any behavior is part of the transaction and participation in economic activities. The transaction accounting is realized through the blockchain, which is used for subsequent cross-validation and behavior traceability. And contribution statistics to provide the basis. This is similar to recording everything you do on the ledger.

The second is the payment tool between machines . Inter-machine payment can be through financial accounts (including bank deposit accounts and payment institution accounts) , or through Tokens in the blockchain. When the number of nodes in the machine network is large, the efficiency of payment through financial accounts may not keep up with the efficiency of interaction between machines. In addition, the pre-approval requirements of financial accounts are difficult to be compatible with the openness of the machine network. The payment through the Token in the blockchain is essentially to make the blockchain a value settlement protocol in the machine network. In this respect, it is necessary to face the network effect of currency and the problem of currency mismatch.

Machine networks also exist in a world dominated by fiat currencies . For example, the capital expenditures of the Bitcoin network (mainly mining machine R & D and acquisition costs, etc.) and operating expenditures (mainly electricity costs and personnel salaries, etc.) are denominated in fiat currency. Fiat currency has a strong network effect. If the machine network uses Token that is not linked to fiat currency as a payment tool, there will be a currency mismatch problem. For example, Bitcoin miners need to bear the risk of Bitcoin price fluctuations against the US dollar.

From the perspective of economic mechanism design, the machine should be given a stable incentive to ensure that economic activities in the machine network are not disturbed by fluctuations in the value of payment instruments. Providing a suitable inter-machine payment tool in the central bank digital currency and the global stable currency (represented by Libra) is the basis for introducing financial activities in the machine network. In turn, the huge number of nodes in the machine network and the rich economic interaction between the nodes also provide a good scenario for the promotion of central bank digital currency and global stable currency applications. In this link, a mapping relationship between the machine's central bank digital currency / global stable currency wallet address and the machine ID is required.

The third is the economic model design of the machine network. Although a single machine has a controller or owner, the machine network as a network has no owner and has strong public product attributes. In the early stage of development, any network is small-scale, and the network effect is very small. New entrants need to pay for switching over and bear early risks. How to motivate new entrants to make them compatible with the further development of the network? In this respect, a distributed economic model and a floating value token should be introduced. Through programmability design, the token can be effectively captured from economic activities in the machine network. Only in this way will the machine network have endogenous growth power. Starting from programmability, I discussed how blockchain can become a large-scale value settlement protocol .

"Machine network + blockchain" will generate great economic value. First, provide a new source of income for the builders, operators and maintainers of machine networks. In particular, the realization of a fair and effective economic model-all machine nodes that contribute to the network should be rewarded. Second, develop new financing methods for network infrastructure construction. Third, it truly brings the blockchain into the broad application scenario of " Internet of Everything ". We must have full imagination for these new economic forms that will emerge.

Large-scale collaboration between machines is not fantasy

Back to the question raised at the beginning, my answer is: by introducing a market mechanism in the machine network, a decentralized, value-exchange-based large-scale collaboration mechanism between machines can be realized, allowing machines to collaborate like humans . The Bitcoin network is a prototype system of "machine network + market mechanism". The core of "Machine Network + Market Mechanism" is the economics of " Machine Network + Blockchain ". I discussed how to express the machine network as an economic problem through the "two-step method", as well as three important economic problems: the combination of machine ID and address in the blockchain, the payment tool between machines, and the economic model of the machine network design. These issues are also very important for understanding the future machine economy and equipment democracy.

Large-scale collaboration between the blockchain and the machine will generate great economic value, develop new financing methods for the construction of network infrastructure, and bring the blockchain into the broad application scenario of " Internet of Everything ". Dr. Xiao Feng pointed out that the blockchain facilitates us to use smart contracts to automate value measurement, value distribution, value storage, and value settlement for large-scale collaboration, making cross-border, public affairs large-scale collaboration feasible, credible, and Be efficient. Large-scale collaboration between blockchain and machines will be an important issue in this regard.