SACT-AI’s Reinforcement Learning for Realization of Keynes’s Bancor/ICU Proposal at Bretton Woods
Posted on | August 10, 2026 | No Comments
Citation APA (7th Edition)
Pennings, A.J. (2026, Aug 09) SACT-AI’s Reinforcement Learning for Realization of Keynes’s Bancor/ICU Proposal at Bretton Woods. apennings.com https://apennings.com/artificial-intelligence/sact-ais-reinforcement-learning-for-realization-of-keyness-bancor-icu-proposal-at-bretton-woods/
Introduction
Keynes’s monetary proposal for the Bancor and International Clearing Union at Bretton Woods was rejected for a US dollar tied to gold. The US had the military power, the economic productivity, and the financial capability to enforce a new global currency and trade order. Yet the rejection was also technological: the 1944 tabulating machines, punch cards, and high-latency telegraphy/radiotelephony simply could not execute the dream of the dynamic, symmetric, rule-based coordination Keynes required.
Can the SACT-AI Bancor/ICU system, reconfigured with reinforcement learning, finally supply the missing computational substrate for global currency coordination? Within this proposed architecture, Q-learning emerges as the primary policy-optimization engine, with PPO (Proximal Policy Optimization) serving as a complementary fine-tuning layer. Together they overcome the 1944 technological constraints while maximizing transitional USD influence and neutralizing the Milkshake Theory’s extractive suction.[1]
Keynes’s Original Bancor/ICU Proposal and the 1944 Technological Barrier
Keynes’s 1943 White Paper envisioned the ICU as a supranational ledger operating through Bancor, a neutral international bank-money created and extinguished solely through book entries. Core mechanics included:
– Quotas based on pre-war trade averages functioning as automatic overdraft facilities (free up to 50 %, modest interest to 100 %, escalating penalties beyond).
– Multilateral netting of all bilateral claims, eliminating physical settlements.
– Symmetric penalties on persistent surpluses (1 % on excess above 50 % of quota, rising to 2 % above 100 %), with proceeds recycled automatically as new overdrafts to deficit nations.
– Tier-calibrated adjustment rules that specifiy deficit nations could devalue or draw facilities, while surplus nations faced revaluation or lending obligations (Keynes, 1943; Piffaretti, 2009; Whyman, 2014).
These rules required continuous, real-time monitoring of trade balances, reserve positions, and capital flows across dozens of nations — followed by instantaneous ledger updates and policy adjustments. In 1944, the available technology made this impossible. Tabulating machines (IBM Hollerith-derived systems) processed data only in rigid batch runs using physical punch cards (80-column format) at speeds of 100–300 cards per minute. Operations were strictly sequential: cards had to be physically transported, sorted, and collated by armies of operators. There was no stored-program capability, no random-access memory, and no capacity for conditional branching or real-time updates.
International connectivity relied on telegraphy and early radiotelephony, with latency measured in hours or days and no packet-switched error correction. A single multilateral clearing run or quota adjustment across 44 nations would have taken weeks of manual reconciliation, rendering Keynes’s automatic, symmetric mechanism operationally unfeasible (IBM History, n.d.; Gladwin, 2000). The Bretton Woods dollar-gold standard prevailed precisely because it was the only ledger simple enough to be computed with 1944 tools.[2]
Q-learning is the primary policy-optimization engine for a SACT-AI Bancor/ICU because the problem is best framed as learning the value of discrete, rule-based adjustment actions in a high-stakes, partially observable environment with abundant historical data. PPO offers superior stability and is better suited for continuous control sub-problems, but its on-policy nature and lower sample efficiency make it secondary.
In Global Spreadsheet Capitalism, the combination of Q-learning’s value-driven symmetry with PPO’s fine-tuning capability creates the most robust, auditable, and adaptive engine for realizing Keynes’s vision of a computable, multipolar monetary order.
References
Gladwin, J. (2000). The IBM Hollerith tabulating machines. IBM Archives.
IBM. (n.d.). The Punched Card Tabulator. IBM History. https://www.ibm.com/history/punched-card-tabulator
Keynes, J. M. (1943). Proposals for an international clearing union (Cmd. 6437). His Majesty’s Stationery Office. (Reproduced in IMF eLibrary, 2010). https://www.elibrary.imf.org/display/book/9781451972511/ch001.xml
Lenzu, S. (2026). Artificial intelligence and monetary policy: A framework and perspective on cyclical transmission, structural transition, and financial stability. Federal Reserve Bank of New York & NYU Stern.
Piffaretti, N. F. (2009). Reshaping the international monetary architecture: Lessons from Keynes’ plan (Policy Research Working Paper No. 5034). World Bank.
Whyman, P. B. (2014). Keynes and the International Clearing Union. In The political economy of the Keynesian revolution (pp. 1–31). Lancashire University.
Notes
[1] This project extends my Master thesis on the deregulation of finance and telecom in the 1980s looking at the emergence of SWIFT and Reuters Monitor and Dealing currency and news terminals after the breakdown of the Bretton Woods system in the 1970s.
[2] My early work on the “Smith Effect” shows Adam Smith’s political economy had a major influence on the development of information technologies through its influence on moving the conception of weath from the treasury of the sovereign to the exertions of the population, leading to the importance of Herman Hollerith’s Census Machine and the eventual development of International Business Machines (IBM).
AI Prompt(s) Describe why Q-learning would serves as the primary policy-optimization engine in a SACT-AI Bancor/ICU System. Connect the last two answers more closely with Keynes’s orginal Bancor/ICU proposal at Bretton Woods and the tabulating and telegraph technology of 1944.
© ALL RIGHTS RESERVED
Not to be considered financial advice. AI is often used, and results are thoroughly interrogated. Links are used for some citations.
Anthony J. Pennings, PhD is a Professor at the Department of Technology and Society, State University of New York, Korea and a Research Professor for Stony Brook University. He teaches AI and broadband policy. From 2002-2012 he taught digital economics and information systems management at New York University. He also taught in the Digital Media MBA at St. Edwards University in Austin, Texas, where he lives when not in Korea.
Tags: Bancor/International Clearing Union (ICU) > John Maynard Keynes > PPO (Proximal Policy Optimization) > punch-card tabulating machines > Q-Learning > reinforcement learning (RL) > Transatlantic Telegraphy
Comments
Categories
Referencing this Material
About Me
Professor (full) at State University of New York (SUNY) Korea since 2016. Research Professor for Stony Brook University. Moved to Austin, Texas in August 2012 to join the Digital Media Management program at St. Edwards University. Spent the previous decade on the faculty at New York University teaching and researching information systems, digital economics, and global political economy
You can reach me at:
anthony.pennings@gmail.com
apennings70@gmail.com
anthony.pennings@sunykorea.ac.kr
About me
Writings by Category
- artificial intelligence
- automatrix
- blockchain
- books and films
- characteristics of digital media
- creative cities
- creative industries
- crisis communications
- cryptocurrencies
- cultural Industries
- data analytics and meaning
- democratic political economies
- digital coordination
- digital destruction
- digital geography
- digital media economics
- digital media management
- digital monetarism
- digital spreadsheets
- disaster risk management
- disruption
- dystopian economies
- educational innovations
- electric money
- enabling frameworks
- enterprise systems
- financial news
- financial technology
- food policy and security
- futurists
- geopolitical risk
- global communications
- global e-commerce
- health IT
- how IT came to rule the world
- intellectual property
- journalism
- meaning makers
- meaningful play
- media strategies
- mobile technologies
- multimedia
- political economies in sf
- political economy of media
- public administration
- science and technology studies
- sensing technologies
- smart new deal
- social media
- space systems
- spreadsheet logic
- strategic communications
- sustainable development
- technologies of meaning
- technology management
- telecom policy and net neutrality
- telegraphic political economy
- telelearn/telework
- the cold war
- the smith effect
- uncategorized
- visual literacy and rhetoric
