Self driving cars lack ambition, we need to use machine learning techniques to plan the economy better.
Source: Mike Birdy
Adam Smith was no businessman. A lifelong academic and writer supported financially by wealthy donors, he championed the interests of the prosperous consumer against the tradespeople who laboured to provide meat, beer, and bread. The invisible hand of the free market and free trade was able to provide all the necessaries and luxuries that he might wish to buy.
What was invisible to Adam Smith was the hard, hard work of formulating an idea, selling an idea, producing a business plan, arranging credit, managing employees, dealing with regulations and taxes, the logistics of getting goods to market, taking decisions and taking risks about what markets to chase next. Even less visible was the inevitable collateral damage of unsold goods, excessive production destroyed the ability to maintain prices, shipwrecked endeavors, failed entrepreneurs, debtors were imprisoned, bailiffs enriched, families desperate, and widows starving. In other words, the whole process was and is desperately wasteful of resources, wasteful of human capabilities and discouraging to genuine progress.
A view devised by Nobel laureate in economics Robert Schiller to account for the ‘secular stagnation’ of the economy is that people are now fearful of spending and borrowing because they believe that their jobs may soon be lost to further automation by machine intelligence. According to the American Trucker Association, there are now 3.5 million professional truck drivers in the USA. Self-driving vehicles are the first target of the new artificial intelligence industry, riding on work done with drones and robotics for the defence industry. Their widespread adoption is almost an economic inevitability.
The spectre of mass job losses and spreading poverty haunts the imaginations of the most thoughtful people in every business today. Doing nothing is probably not going to be an option this time around. But the advances in the capabilities of machine intelligence open up the possibility of applying it to just these kinds of economic problems. Machine intelligence may just be smart enough to help solve itself. Innovations in three areas have made this possible. Business standardisation, computing technology, and legal innovations.
Transparent and globally accepted International Financial Reporting Standards (IFRS) are being matured and rolled out. Business IT systems have become standardised using products like SAP and Microsoft Dynamics. Almost every significant financial transaction is recorded somewhere electronically. The Internet has made it possible to collect and store very large datasets of information. These datasets, along with the processing power of graphics cards designed originally for the computer games industry have made progress in machine learning and artificial intelligence possible where it had stalled in the past. Cloud based software and hardware from companies like Amazon Web Services have removed cost barriers to setting up powerful computer systems. These technical and business advances, together with the legal innovations like Creative Commons and the free availability of open source software libraries bring even very hard problems like the economy within reach of research and solutions.
Back in the classroom, business acumen is developed in MBA classes often using large data-based case studies to develop experience and insights. The complexities of business situations and decisions are revealed in the study of key cases and comparative case studies. But there is no real substitute for learning from experience. But using large amounts of data to infer patterns, learn from examples and to classify and formulate responses is exactly where machine learning excels. With the ready availability of processing power, standardised data inputs and removal of cost barriers, the automated control of businesses processes and business administration and ultimately the economy as a whole could be as achievable as the automated control of road logistics.
It may even prove to be the simpler problem.
Work has already been done in conceptualising the economy as a cybernetic system, but perhaps understandably not in the United States. This work was mainly done in the then Warsaw Pact country of Poland, using exactly the same mathematical concepts and techniques that form the basis of machine learning. Oskar Lange at the University of Warsaw was the mathematician and economist behind this work, published in his Introduction to Economic Cybernetics in 1970. His purpose was to develop a more effective and centrally planned socialist economy. The Alfred Nobel University, in Dnipropetrovs’k, Ukraine still offers a course in Economic Cybernetics but, alas, the course content now seems typical of computer engineering courses worldwide.
However, the reflexive anti-communism that gave is the Vietnam war, the McCarthy era and the doctrine of Mutually Assured Destruction should not prevent us from understanding and using these insights for the benefit of the country’s economy and people. In ‘The State and Revolution’ Lenin opines that ‘Accounting and control–that is mainly what is needed for the “smooth working”, for the proper functioning, of the first phase of communist society.’ Maybe machine learning can help with that. Maybe also, who knows, the efficiency savings from cybernetic business administration may even be enough to pay the cost of a universal basic income – once the machines take over all the jobs.
By Ron Ellis