COLUMN-AI will replace most humans, but then what?: Jen

By Stephen Jen

LONDON, August 19 (Reuters) - Is technology more job augmenting or job replacing? This has been a long-standing debate. But recent academic work suggests that technology has been a net destroyer of jobs for decades.

Artificial intelligence and robotics could rapidly accelerate this trend, with significant implications for inflation, the size of government and U.S.-China relations.

Over the long arc of history, technological advances have enabled industries to emerge, as workers, released from "older" jobs by machines, have been able to transition into newer ones.

Indeed, 60% of workers today are employed in occupations that did not exist in 1940, or 74 percent if we consider just the professional category, which added the most workers during the past eight decades.

However, recent academic research suggests we may have reached an inflection point in the U.S., whereby technology is now destroying more jobs than it is creating.

David Autor, an economist at MIT and winner of the 2005 John Clark Bates Medal, argues that since 1980, the jobs replaced by automation have not been fully offset by new jobs created.

This reflects the pace of technological change and the fact that advancements are now increasingly focused on “professional, technical, and managerial occupations,” Autor notes, rather than lower-skilled work.

He finds that machines that are more powerful than an average human (e.g., a tractor) are typically labour-augmenting and productivity-enhancing, while machines that are also smarter than the average human tend to be labour-substituting.

And AI is on pace to be a lot smarter than most humans.

While forms of AI have been around since the 1940s, the immense computing power resulting from advances in semiconductor technologies has now allowed machines to attain multidimensional intelligence.

It is therefore reasonable to assume that many workers are going to be replaced by automation in the coming decades, even if the best AI is never as creative or imaginative as the smartest humans. In fact, a 2019 OECE report and a 2018 paper by PriceWaterhouseCoopers argue that some 15-30% of all jobs in developed markets are at risk of being automated.

IMPLICATIONS

If AI does turn out to be a net job destroyer, what are some of the biggest implications?

First, it’s likely to be deflationary. High and rising unemployment resulting from ever cheaper and more capable machines should, in theory, lead to structural deflation, as technologies that can rapidly augment the supply of goods and services should reduce demand if they cause massive job losses.

Next, the U.S government will probably get even bigger. In a mass unemployment scenario, the government would likely be compelled to step in to facilitate income and wealth transfers from the owners of robots and tech businesses to the unemployed workers.

And which countries will come out on top? The economic winners and losers in the years ahead will likely be determined by who can best create and utilize technology.

The U.S. and China, both dominant in cerebral technologies, therefore appear well positioned to thrive in this environment. These economic and technology superpowers have adopted muscular industrial policies, while Europe – the other big regional power – has not yet done so.

What this also suggests is that, even if the trade war between the U.S. and China is short, the tech war between these two countries could be protracted – and ultimately much more consequential.

The tech war, unlike the trade war, is dynamic, meaning it’s not about challenging the static comparative advantages of nations, but rather continually evolving and advancing.

Investors would be wise to keep this distinction in mind, as the dynamic aspect of the tech war is apt to become much more important than, say, whether Vietnam is allowed to sell cheap running shoes in the U.S.

EXPONENTIAL CHANGE

My views here are admittedly speculative. But the arguments for why AI and robotics could ultimately be labour-creating are as well.

Furthermore, these arguments are often obscured by sloppy references to labour productivity, which is a simple ratio between output and labour input.

When calculating this, there is often little explanation of what part of the output should be attributed to the labour input. For example, should subway train drivers account for the value of the entire subway system? Projections based on such questionable assumptions should be viewed cautiously.

Finally, it's also true that populations in many developed markets are aging, so the heavy use of automation could simply offset the shrinkage in the labour force, something we’re already seeing in Japan and South Korea.

But aging, like natural evolution in general, is gradual, while computational and technological evolution accelerates at an exponential pace. Because of the convexity in technological advances, it’s hard not to bet on technology rather than workers.

(The views expressed here are those of Stephen Jen, the CEO and co-CIO of Eurizon SLJ asset management).

Enjoying this column? Check out Reuters Open Interest (ROI), your essential new source for global financial commentary. ROI delivers thought-provoking, data-driven analysis of everything from swap rates to soybeans. Markets are moving faster than ever. ROI, can help you keep up. Follow ROI on LinkedIn, and X.