Editor's note
This article is authorized to be reproduced from Peking University Innovation Comment, and the content comes from the 2022 INNO CHINA China Industry Innovation Conference-Financial Technology Innovation Forum, Professor Deng Xiaotie 's report shared by Peking University Frontier Computing Research Center .
Character introduction
Deng Xiaotie is currently the chair professor at the Center for Frontier Computing Research at Peking University, the first director of the CCF Computing Economics Professional Group, the director of the CSIAM Blockchain Committee, and the director of the Multi-Agility Center of the Institute of Artificial Intelligence of Peking University. His main scientific research directions are algorithm game theory, blockchain, Internet economy, online algorithms and parallel computing. In 2008, he was elected as an ACM Fellow for his contribution to the field of algorithmic game theory; in 2019, he was elected as an IEEE fellow for his contribution to the field of incomplete information computing and interactive environment computing; in 2020 he was elected as a foreign academician of the European Academy of Sciences; in 2021 he was elected as a CSIAM Fellow of the Chinese Society of Industrial and Applied Mathematics (CSIAM Fellow); in 2021 he was appointed as a director of the Game Theory Society (GTS); in 2021 he was appointed as an honorary director of the Game Theory Branch of the China Operations Research Society; in 2021 he was awarded the CCF Artificial Intelligence Society Multi-Agency and Multi-Agency System Research Achievement Award; in 2021 he was awarded the "Test of Time Award" for ACM Computational Economics.
01
Research and development of computational economics theory
A key in computational economics is equilibrium. The most intuitive feeling of economic equilibrium is that merchants and users in the market achieve through consumption and supply behaviors, and the market can be derived from equilibrium price to market clearance. There are a series of technologies for clearing the market, starting from the "invisible hand" proposed by Adam Smith) to Walras Price (Walras's generally has a equilibrium price. This view believes that the market can be cleared by just adjusting the price. This is a good technology in terms of calculation, although the stage of the theory proposed at that time could not achieve this) , and finally the solution to the equilibrium price that can be seen from the perspective of computational economics.
In terms of calculation, the initial calculation of Walras Pricing is trial and error. The human economic society has been in a equilibrium state for a long time, and this method is suitable for slow-changing economic society. A major progress was proposed by Langer (Oskar Ryszard Lange, Polish economist) . He talked about the problem of computational economics from the perspective of planned economy, and believed that through computer simulation, the state reached by market economy can be achieved, including social optimization (social optimization) and social benefits (efficiency) , but it is difficult to apply to large-scale economic society. The founder of mathematical planning, , George Bernard Danzig, (George Bernard Dantzig) proposed to use PILOT Model to analyze the role of energy in the economy, which is also implemented from the perspective of equilibrium computing.
When reviewing this set of ideas in a whole, it can be defined very clearly from the calculation perspective: that is, PPAD (polynominal parity argument directed, polynomial time directed graph parity argument) . This type of problem solving algorithm is equivalent to fixed point calculation.
02
fintech important practice
digital cryptocurrency design for blockchain
fintech now has the main set of method practices, which is the design of digital cryptocurrency for blockchain. The basic principle seems to be contrary to the development route of financial economics : it increases a lot of friction and fixes the supply. The production of its value is supported by the market price of electricity consumed by miners. Bitcoin’s design is very mechanized and equilibrium is not easy to achieve. It can be noted that the changes in Bitcoin price are not like an equilibrium market.Its balanced and stable implementation depends on the joint participation (investment) of a large number of miners (computing power).
Digital Economy Game Financial Basics
The meta-universe constructed before Bitcoin emerged, the second life (Second Life) The financial foundation of the digital economy game is a good example. It hopes to maintain the exchange rate of fiat currency (US$) and its virtual currency (Linden Dollar L$) exchange rate . Over the past 10 years, its redemption rate has been stable between L$270/US$1 and L$240/US$1. It is a virtual economic society that simulates the real economy and is implemented from the central system architecture of Internet 2.0. And that's not the case with Bitcoin. The key market problem of the digital economy and the metaverse is how to establish a market equilibrium price. In the world of cryptography currency, this has never been solved.
From a computing perspective, humans have limited rationality in computing and may not find the optimal one. Taking the economy of Bitcoin as an example, it has undergone a lot of simplifications through calculations. For example, production volume is pre-specified, production cost is determined by the energy consumption involved in production (PoW), and product consumption is determined by the use of cryptocurrency. It uses a "PoW fixed consensus time" to generate a block in about 10 minutes. Under the framework of Bitcoin, observing the pricing in establishing an NFT market economy is actually as unstable as the pricing of the market economy itself.
There is actually a big difficulty in digital economics lies in the difficulty of calculating equilibrium. Because many things inside are not natural and can be twinned, but a big factor is the digital economy. From the perspective of computational economics, computational equilibrium is a problem of difficulty in PPAD. PPAD mentioned earlier that it is equivalent to Nash equilibrium, equivalent to market equilibrium, and equivalent to fixed point calculation. In a dynamic state, it is equivalent to the equilibrium of behavioral strategies in the Markov game process.
talks about the game theory challenges in digital society. We ask a few questions casually, such as how to determine the equilibrium price of NFT? How to connect the real economy and the digital economy? Let’s talk about the real economy here. Because its price changes are relatively long, can a market that can achieve equilibrium and a market that cannot achieve equilibrium achieve mixed equilibrium? Fiat currency and digital currency Pegging How to solve the equilibrium calculation of digital society? Regarding this, there are many known technologies in the system of computational economics. One of them is the finite rational proposed by Herbert Alexander Simon (Herbert Alexander Simon) . There is an approximate calculation solution in time, and there is competition in information compared to (Competitive ratio) . There is an government cost in social welfare issues, and there is an incentive compared to (Incentive Ratio) . So how to achieve the simplification of the problem?
03
Innovation key methodology: Game under private information
Game theory In order to overcome the equilibrium non-existence of pure strategy Nash equilibrium (Pure Strategy) has been introduced. Equilibrium in economics and finance both relies heavily on the introduction of such a new concept of solution: a hybrid strategy based on probability theory. There may be differences in how to understand probability theory in reality. It is also possible to be manipulated by opponents described by probability theory.
The majority vote designed in Bitcoin decisions can ensure balanced implementation in this environment, and the consensus that the expectation of (common knowledge) will be destroyed. Here we can build it with an Markov Decision Process model (Markov Decision Process, MDP) . Some people can destroy the consensus of Bitcoin, and this group can be destroyed when it is less than 50% of the total number of people.
The key to this is called selfish mining is the application based on private information. This small group can hide its work, which is inevitable in the digital economy today.Privacy protection even protects this legally. Therefore, the stability of the entire Bitcoin is far from reaching the level that we can grasp in the digital world. Further, we can see that there can be more complex strategic behaviors. There is a theory here called cognitive hierarchy (cognitive hierarchy) . Add to that another group of higher-level people who can mine with vision and profit from higher-level cognitive strategies.
In the digital world, price war (Price War) can be established very accurately. This is a computational model that can be used to provide subsidies to consumers during holidays. Here we have a lot of key information. Unlike blockchain, we don’t know the opponent, the opponent’s strategy, the opponent’s value function, we still have imperfect information, and we don’t know the opponent’s historical behavior. Here we have introduced another financial technology that uses a lot of "deep learning cognitive decision-making".
04
AI Industrial application of economics
can generate "automatically generated design" in financial technology, which cannot be done in traditional economics. For example, the consensus mechanism of PoS (proof of stake) in blockchain can be achieved through many techniques to achieve automated decision-making frameworks, which is also a special feature of blockchain.
Compare traditional economics with AI economics. In traditional economic pricing, we consider factors such as static world, market equilibrium price, auction pricing mechanism, fixed price and other factors. In AI economics, we can build a multi-agent learning process to discover unknown content, and we can build a framework of nearly perfect information. AI economics is a very interesting technology recently.
Back to the auction issue, the very famous theory in the auction is called Myerson's seller's best-in-return auction plan (Myerson's auction) . It can design a quotation strategy so that auctioneers can get the highest profit. Here is a prior calculation (prior). We know everyone's price function, and our understanding of it is a common knowledge. At this point, there have also been a lot of very interesting work starting from the machine learning method of and deep learning method. Here we mention David C. Parkes, a major contributor to AI economics, and we can see that the machine design we talked about in economics before can achieve this with deep learning methods.
05
Breakthrough point of financial technology innovation
Finally, we come back to a question, which is the financial challenges faced by computational economics. Let’s take a look at the Financial Stability Board, which is an American institution. They have discussions on the design and game of many international, especially cross-border finance. It mentions what impact FinTech will have on our world in the future. It mentions here identify regulatory and supervisory issues raised by FinTech, allowing us to prevent its impact on financial stability.
Finally, looking back at the history of human development, from gold and gems to later printing presses and banknotes, to today's credit cards and electronic payments, there are various ways to make people spend money. For computational economics, there is a very important challenge here, that is, when building each financial system, can we establish a set of technologies for its equilibrium calculation to complete the analysis of the stability of the newly built financial system (stability) .
CCF Computational Economics Professional Group Introduction:
Computational Economics Professional Group was established on August 7, 2022 at the Chinese Computer Society (CCF) Suzhou Business Headquarters. It was established on August 7, 2022. It coincided with the 60th anniversary of the establishment of CCF. The Computational Economics Professional Group is the 40th professional group of CCF and is also the first professional group with interdisciplinary integration characteristics.The professional group was initiated by Professor Deng Xiaotie from the Frontier Computing Center of Peking University and Professor Dong Zhiyong from the School of Economics of Peking University, as well as more than 200 experts, scholars and corporate leaders, and Professor Deng Xiaotie serves as the founding director. The development planning direction of the
professional group includes: promoting the intersection of computer and economics, cultivating talents in the computing economy direction, strengthening the contact between professional groups and government departments, organizing professional groups to cooperate with enterprises, and leading the collaborative thinking tank platform construction. Research directions include: basic theory of computing economy, platform economy , digital economy, intelligent governance and optimal policy design, etc. The professional group is committed to connecting and uniting the vast number of researchers in this field, organizing academic activities, enhancing academic exchanges, and promoting the development of research and application in the interdisciplinary fields of computing economy. Under the leadership of the CCF, the professional group assists the computer field in the formulation of national policies, widely participates in economic policy discussions and formulation, and collaborates with other CCF branches to build a national computing economic think tank platform.
edit | Alexandra
source | Peking University Innovation Review
