Wild Claims Today
AI will take over the universe. At a minimum it will eliminate 10% of jobs, really important skilled jobs, leading to a downward spiral of joblessness and economic decline. The experts say so. AI has clearly shown exponential growth in performance. Do the math. It is unavoidable.
[Amateur Human] Analysis
We have experienced so many “revolutions” that we can’t keep track of them. Electricity, railroads, telegraph, radio, tv, Marxism, existentialism, globalism, free trade, colonialism, skepticism, environmentalism, feminism, Freudianism, behavioralism, racial equality, women’s rights, gay rights, sexual experience, cold war, energy crises, terrorism, modern finance, lean six sigma, meritocracy, media, computer, communications, cell phone, biotechnology … Few have had material long-term impacts upon national or global economies.
Economies today are comprised of components. They are not just farming. They are not just basic manufacturing. Financial and real. Raw materials, component suppliers, manufacturers, logistics, wholesalers, retailers, installers, returns. Various industries. Various functions. Goods and services. Domestic and international. Importers and exporters. This complexity means that no one technological, process, regulation or other change is likely to have a HUGE impact.
Labor is only one input component. It combines with land/natural resources, capital/facilities, technologies/processes and financial resources to produce outputs. Total compensation is 54% of the US economy. Even the elimination of 10% of all jobs would only result in a 5.4% reduction in the total costs of production. In the short run, this would disrupt labor markets. In the medium-run, profits and rents would increase from the cost savings. The loss of incomes would reduce demand for goods and services, reducing profits, rents and compensation. Disruption, yes. Disaster, no.
The US economy today is less subject to the high fixed cost operating leverage that drives business cycle booms and busts. This reduces the frequency and depth of depressions. It improves the US economy’s ability to recover from exogenous shocks like an oil crisis, mortgage meltdown, pandemic or possible decrease in labor demand.
AI requires a human to initiate, fine-tune, launch and implement it. Only a small slice of the workforce can use AI tools. These individuals will be cautious when applying tools that threaten their personal employment.
Historically, “analysts” have employed newly available productivity improvement tools. There are only so many analysts with available time to evaluate and apply AI tools directly.
We have a history of analysts/STEM/creative class professionals adopting personal productivity tools. None of these situations resulted in significant/quick reduction in the demand for labor. They made these individuals more productive leading to the continued growth in these job categories for 40 years. Consider spreadsheets, query tools, report writers, database management software, ERP systems, statistics packages, CAD/CAM, engineering graphics workstations, desktop publishing, simulation software, project management software, warehouse management systems, EDI, internet search tools, forecasting software, shop floor modeling, activity-based costing, balanced scorecard, credit analysis tools, insurance underwriting tools, marketing research/survey tools, product development tools, logistics optimization tools, event planning and management tools, reverse logistics software, repair depot software, agriculture crop planning, the list is endless.
IT and process improvement projects are well structured today. These project teams will evaluate all dimensions of AI technology, considering direct and indirect costs, benefits, risks and off-ramps. They will act as a brake on the speed of adoption.
The jobs where “thinking” is the critical/90%+ requirement are few. Most jobs have thinking, feeling and doing dimensions. They employ local workers. Most “thinking” jobs combine inductive and deductive reasoning, multiple intelligences, hard and soft skills. AI focuses on the pure logical thinking dimension.
Many jobs require analog, travel and emotional physical interaction. AI cannot help.
Most corporate advances reply upon “matrix management” and project team skills. AI does not fit easily except as a tool for solving technical problems.
There are some research, logistics, analysis and customer service jobs that will be mostly replaced by AI. A human evaluator and quality control will remain.
Some scientific jobs will be quasi-automated, reducing employment.
The adoption rate across industries varies greatly. Manufacturing, distribution, logistics, electronics, IT yes. Others, much more slowly.
The “lump of labor” fallacy must be overcome. Individuals have a variety of employable skills. “Next skill up” applies. Individuals who lose positions find “next best” positions in America’s dynamic labor market. That is, job holders at all levels have packages of general purpose, professional, technical, industry and firm specific skills and experiences. AI tools reduce the opportunities to use some of their skills in some industries and firms. They still retain other marketable skills and experiences. A tight labor market will result in unemployment, some long-term unemployment, some avoidable long-term unemployment and a ripple effect that drops generally higher skilled candidates into previously lower paid roles. There will be a lowering of compensation throughout the labor market.
The “automation” of jobs in the physical, clerical and professional world has proceeded for almost 50 years. Employees have adapted effectively, adding skills or moving industries or locations.
Some positions rely upon “tribal knowledge” or individual/personal relations. They are buffered from AI.
At the macroeconomic level, the substitution of capital (AI computers) for labor is well understood. Firms reduce their costs and improve their competitive positions. They earn greater profits for a while. Slow to adapt firms die. Owners receive the profits. Owners consume more of the goods and services with a high marginal income elasticity. Personal services, professional services, travel, hospitality, leisure, tax, investment, security, charities, and construction grow. Total demand for goods and services remains the same or increases. The adoption of productivity increasing technology increases the supply side (capacity) of the economy. “Say’s law” ensures that in the long run that capacity will be used because the factor returns (rent, labor, profit) create new demand. In the short run we can have recessions and depressions. If we don’t use known fiscal and monetary policies to increase demand to use capacity we can stay in a recession, but we have 75 years of experience in dozens of modern economies to guide us.
This is not intended to diminish the potential role for AI. It will eliminate some jobs and complement others, making them more productive. It will solve problems faster than humans and hence open the door for new goods and services. It may reach superintelligence and seek to take over the world, but I’m not addressing that possibility.
Summary
AI is a tool that can replace some of our existing labor force. It may impact a significant share of the labor force in the next decade. Our economy is large, complex and dynamic. History indicates that some positions will be eliminated but many more will be boosted by the new man/machine capabilities that make them even more productive. There still appears to be no limit to the demands of individuals for goods and services, so it is very likely that this will improve the economy’s productive capacity and that we will consume all of this increased capacity, adding new goods and services to our menu of consumption.
I’m not saying that the labor market, goods and services markets or macroeconomic transitions will be smooth or pain free. Professional workers will have to find new “next best” roles that leverage other new or existing skills. Personal services industries are likely to grow while goods and resources industries decline. Fiscal and monetary policy makers will have to supplement aggregate demand without causing inflation or spooking bond buyers.
The modern advanced economy is truly a miracle. Without central planning or control, it produces the goods and services that are demanded. It adapts to changes in demand, technologies, suppliers, labor markets, resources, finances, regulations, etc. It will effectively adapt to the impact of AI on the demand for labor.
I encourage you to listen to the economics profession on this topic. It is not a new topic. It fits within existing conceptual models and measurements.
IT/AI leaders are not economists. They are subject to the same kinds of irrational thinking that economists have corrected for more than a century. Economic systems seem to be “too good to be true”. There are times when claims are exaggerated for political reasons. Economic systems are not perfect, but their overall functioning is well understood. In theory and in practice, our economies are very well positioned to digest the productivity improvements of AI for the benefit of total capacity and individual consumption.
Good News 