The Techno-Pessimists: Same Old, Same Old.
By Alberto Moel, Vice President Strategy and Partnerships, Veo Robotics
Welcome back to yet one more post on the macroeconomic impacts of automation. The debate is far from settled, and as part of the framing, we developed a consult-y 2x2 matrix outlining four possible positions to take in the debate.
In previous blog posts we did a goofy take on pop culture automation and outlined the case of the techno-optimists. Now, we’re focusing on the views of the techno-pessimists, who predict that AI is likely over-hyped and that we're only going to see "more of the same": continued weak employment growth without increased labor productivity.
The premier proponent of the techno-pessimists is Northwestern University professor Robert Gordon, who makes a compelling case in his monumental 784-page work, "The Rise and Fall of American Growth: The US Standard of Living Since the Civil War." The arguments set forth below are pretty much based on Gordon's view point.1
Professor Gordon notes that the disappearance of good mid-level manufacturing and service jobs in the US that have been either outsourced or lost to automation, has been accompanied by disappointing productivity growth and rising inequality.
His approach to understanding (but not solving) this paradox relies on the classical economist's "growth accounting" toolbox, first developed by Robert Solow in the 1950s. The unit of analysis in growth accounting is the growth in labor productivity, defined as the amount of real GDP produced from one hour of labor. An increase in labor productivity can come from:
- Increases in the quality of labor, usually proxied by educational levels;
- Increases in the quantity of capital relative to labor (from say, increased automation);
- Increases in the quality of capital, for example, from better or more sophisticated equipment or software tools; or
- An increase in the residual not explained by the above three factors, called "total factor productivity," or TFP. The TFP is also known as "the measure of our ignorance” and is a metric that captures the interaction between the three other components of labor productivity.
Professor Gordon correctly homes in on TFP as being the key to the puzzle of low productivity growth and rising inequality. The other drivers of labor productivity are exogenous and will vary with hours worked, investment rates, technology evolution, educational levels, and the business cycle. Only TFP growth is a clear indicator that labor productivity is rising permanently and independently of its drivers.
And what Professor Gordon finds is not encouraging. Figure 1 replicates Figure 17-2 from his book, capturing TFP growth over specific periods of time.
The great surge in the level of TFP between 1920 and 1970 was the result of the full diffusion of the Second Industrial Revolution’s general purpose technologies (GPTs): electricity and the internal combustion engine.2 There was an incremental bump in TFP growth in 1994–2004 from the delayed impact of the diffusion of computers and digitalization. However, the Second Industrial Revolution created a surge in productivity growth that lasted for 50 years, while the advent of widespread computing had a weaker and more delayed effect, lasting only about 10 years.3
Professor Gordon makes the case that 1920–1970 (and even 1994–2004) is an anomaly, and the period 2004–2014 is indicative of what the future brings. In particular, he makes the point that the 1994–2004 bump in TFP was driven by "an unprecedented and never-repeated rate of decline in the price of computer speed and memory and a never-since-matched surge in the share of GDP devoted to investment in information and communication technology (ICT)."
The flaw in this argument, of course, is that just because those circumstances were unprecedented and haven't been repeated since 2004, it doesn't mean they won't be repeated in the future. You can look up Thomas J. Watson’s or even Bill Gates’ own pronouncements on how large the computer market could be.
Professor Gordon does not dispute the frenetic pace of innovation in digital technology, robots, and AI. However, he notes that the fast pace of innovation doesn't necessarily mean these innovations are having an impact on the growth rates of labor productivity and TFP.
Professor Gordon finds the techno-optimists’ excitement regarding advanced automation technologies, machine learning, and AI wildly overdone, as he points to the slow ongoing growth of TFP over the last decade.
He also shoots down the dystopians (which we’ll dissect further in a future post), rightly pointing out that, "now that the American economy has arrived back at a state of relatively full employment, it is hard to maintain the case that robots and AI are creating a new class of the permanently unemployed."
Professor Gordon's argument is that AI and automation aren’t all they’re cracked up to be, and they pale in comparison to the transformative impact of the previous major GPTs, electricity and the internal combustion engine:
Besides the ATM, the other robot I occasionally encounter is the automated e-kiosk check-in machine at airports; this innovation was rolled out between 2001 and 2005 and has thinned the ranks of airport ticket personnel, just as earlier airline web sites largely displaced travel agents and airline telephone agents; yet the rest of the employees needed to run an airline are still there, including skycaps, baggage handlers, flight attendants, pilots, air traffic controllers, and gate agents.
He further elaborates:
Goods are still placed in supermarket shelves by store employees and by the drivers of delivery trucks for beer, bread, and soft drinks. Checkout lanes at retail markets are still manned by clerks rather than robots, and self-checkout lanes are few and far between. Haircuts, massages, and manicures are still exclusively the province of human workers, as are restaurants with their cooks and wait staff. Hotels still have front desk personnel, and if they offer room service, it is delivered by humans rather than robots. Far from occurring overnight, the shift to robots and job destroying artificial intelligence is occurring at glacial speed.
If AI (in all of its forms) is indeed a transformational GPT, how come we aren't seeing its impact on the productivity statistics? Jobs have been lost to automation for a couple of centuries now, and to computers for a couple of decades. But why do employment and job growth remain robust? Are we just early? Do we have to wait a bit to see the 21st century GPT kick in with force?
Or, is AI not a GPT? That's the case Professor Gordon makes, which by implication means that we have nothing to worry about as AI and AI-driven automation take over our jobs. The issue is that this view also asserts that AI and automation won’t significantly improve productivity or create new jobs. So it’s all just plain hype, and we’re stuck in some permanent post-industrial low-productivity plateau.
To recap: Techno-pessimists like Professor Gordon dismiss the impact of AI and automation technology and don’t expect it to make a lasting difference on our economy. Techno-optimists embrace AI and automation as a GPT and believe it will be good for employment and productivity growth.
Traditionalists, as we’ll see in our next post, also believe AI and automation are a GPT, but they are fearful of it. They expect it to have a negative impact on jobs and job creation, leading to long-term shifts in the nature of employment. In other words, traditionalists don't believe AI and automation will expand the pie, only that it will change the width of the slices.
1 I read this weighty tome so that you, our esteemed reader, don’t have to. I thought it was very good, although afflicted by a serious case of historical extrapolation to future dynamics. Professor Gordon spends hundreds of pages talking about seriously disruptive technology s-curves, and then concludes that since we're at some low-productivity plateau, the long-term future bodes more of the same. This sort of extrapolation could have been applied to the Middle Ages to predict that the Renaissance could never occur. Having said that, it is extremely well-articulated and provides an outstanding survey of technology innovation over the last two centuries. Another, more investor-friendly book in the same vein is Sandy Nairn's "Engines That Move Markets."
2 At an annualized growth rate of 1.89%, TFP rose 2.5x in the period 1920–1970.
3 This delay was well-noted by Robert Solow in 1987 with his classic quip that "you can see the computer age everywhere but in the productivity statistics." It took a while, but they were indeed reflected.