120 likes | 130 Vues
Agents as Methodology: Some Speculations. Rob Axtell. Outline. Topic 1 : Price of computing is falling very (exponentially) fast! Topic 2 : How can we compete with the next generation of ACEs who will have a terabyte of RAM on the laptops?
E N D
AgentsasMethodology:Some Speculations Rob Axtell
Outline • Topic 1: Price of computing is falling very (exponentially) fast! • Topic 2: How can we compete with the next generation of ACEs who will have a terabyte of RAM on the laptops? • Topic 3: If we could build SimEverything would it change anything/everything? • Topic 4: Are there any general principles of economics that are always and everywhere true?
Price of Computing • Facts? • Mathematics and computing as substitutes • Fixed price of mathematics • Declining price of computing • Substitution effect: reasonably clear • Income effect: could computing be a null commodity (neither a ‘good’ nor a ‘bad)? • Will we see an exponential migration to computer utilization and computational expression?
Capital StockTurnover • Given substitution of computation for mathematics, how will human capital stocks respond? • Will there be orderlyturnover or will the overinvestment unwind in a disorderly way? • Could turmoil end in global devaluation of the currency or only local losses?
Moore's Law, I • Mainframes c. 1960 (first microsimulation models): 16 Kilobytes core memory • Desktop microcomputers c. 1990: 16 Megabytes RAM • Laptops c. 2000: 1/2 Gigabyte RAM • Notebooks c. 2010: 16 Gigabytes RAM • Handhelds (?) c. 2020: 1/2 Terabyte RAM? • ??? c. 2030: 16 Terabytes RAM? 1.6 x 1013 bytes/105 bytes/agent = 108 agents
Moore's Law, II • Desktops c. 1990: 16 Megabytes RAM • Desktops c. 2000: 1.6 Gigabyte RAM • Desktops c. 2005: 16 Gigabytes RAM • Desktops c. 2010: 160 Gigabytes RAM? • Desktops c. 2015: 1.6 Terabytes RAM? • Desktops c. 2020: 16 Terabytes RAM? • Desktops c. 2025: 160 Terabytes RAM? • Desktops c. 2030: 1.6 Petabytes RAM? 1.6 x 1015 bytes/106 bytes/agent = 109 agents
How Can We Competewith our future selves? • In my humble opinion (IMHO): • Sugarscape is hopelessly outdated… • …but the Anasazi model will stand for some time • The current generation of artificialfinancialmarkets will be improved upon… • …but soon we will have very realistic simulations • Necessary condition for success: data • Pick problems that can be scaled to available resources
Sim Everything • Say the economics or finance minister of your country came to you and said: “The Americans are building a computer model of the whole world, and we want you to build one for us.” • What would you say or do? • Laugh? • Say you want to hear more? • Commit to doing it but you will need a research team and $10 million/year of research support • plus money for a Summer School!!
Revenge ofHomo Economicus • Imagine a population that wears arbitrary amounts of computing power • What could they do that we cannot? • Get good prices on airlinetickets? • Know exactly when to buy and sell equities? • Figure out what lottery numbers to bet on? • Maybe institutions become more sophisticated and people turn over some autonomy
If everything can be simulated... • …are there really any general economic principles that are universally valid? • ‘law of one price’? • ‘competition picks better products’ • … • Are there any empirical ‘laws’ in economics? • Or, is everything contingent, context dependent, relative and conditional?
Our Non-Elephants • Non-Walrasian markets: • Decentralized exchange (including networks) • ZI exchange • Non-Coasian firms: • Multi-agent firms • Non-Nash game theory: • Adaptive agents, who avoid/eschew fixed points and perpetually seek utility gains
Not the End...but the Beginning Thank you all for a great Summer School!