Martin Wolf talks to David Autor: could AI be a bigger threat to US jobs than China?

Episode Summary

Executive Summary: Martin Wolf and MIT economist David Otto examine the China shock’s long-run damage to U.S. manufacturing communities and argue Trump’s tariffs are the wrong response: they raise costs, weaken export sectors, and don’t restore the old low-wage jobs. The conversation then pivots to AI, which Otto sees as a broader but slower-moving shock that could amplify human capability if designed well, yet risks labor displacement and worse inequality if used mainly to cut costs.

Main Topics: The China shock and local labor-market damage (Priority: 5/5): Otto revisits research showing that the surge in Chinese imports after WTO accession hit manufacturing-intensive U.S. communities hard, producing job losses, lower incomes, and social distress rather than smooth worker reallocation. Why manufacturing shocks hit places, not just workers (Priority: 5/5): The discussion emphasizes that manufacturing is geographically concentrated and often anchored by a few firms or sectors, so closures can devastate entire local economies in ways diffuse job losses do not. Trump tariffs as the wrong answer to a real problem (Priority: 5/5): Otto argues the concerns behind protectionism are understandable, but blanket tariffs are counterproductive: they try to revive obsolete industries, raise input costs, invite retaliation, and undermine future-oriented sectors. China’s shift to higher-tech competition (Priority: 5/5): Rather than wanting to preserve low-value manufacturing, China is now investing heavily in strategic technologies such as robotics, AI, batteries, telecoms, aviation, and quantum computing, creating a new competitive challenge for the U.S. AI as a diffuse, slower-moving shock (Priority: 4/5): Otto distinguishes AI from the China shock: it will affect tasks and occupations across many places, arrive more gradually, and often be seen by firms as a productivity opportunity rather than a direct competitive threat. Designing AI to augment workers rather than replace them (Priority: 4/5): The speakers debate whether AI should be used to remove labor or to improve human judgment in sectors like healthcare, education, law, and skilled repair, with Otto favoring augmentation and broad access to expertise. The political and social stakes of technological change (Priority: 4/5): The conversation closes on concerns that weak institutions, polarized politics, and employer incentives to cut labor costs could produce a dystopian outcome unless policy steers AI investment and adoption carefully.

Key Arguments: The China shock was not just a trade statistic; it produced large local labor-market losses, with unemployment and non-employment rising almost one-for-one with manufacturing declines. Manufacturing losses were especially damaging because manufacturing jobs were concentrated in specific places and often provided relatively well-paid employment for non-college workers. Trump-style tariffs try to recover the wrong kind of manufacturing and ignore that the real contest is over high-tech industries such as AI, robotics, batteries, and telecommunications. Trade restrictions on imports also hurt U.S. producers because many imports are inputs into export sectors; raising those costs weakens competitiveness in autos, electronics, aviation, and telecoms. China’s industrial policy demonstrates that selective protection plus targeted investment can build strategic capabilities; the U.S. needs a comparable innovation strategy, not nostalgia. AI will likely be slower and more diffuse than the China shock, affecting occupations and tasks across the economy rather than wiping out a few local industrial hubs. The best AI policy is to use the technology to amplify human capability—especially in education and healthcare—rather than simply maximizing labor replacement. A major risk is that firms will use AI mainly to lower labor costs, which could intensify inequality and political backlash if workers are not protected and retrained. The U.S. is not running out of work so much as running out of workers, making productivity-enhancing AI and automation potentially valuable if managed well. Future education should focus less on narrow routine skills and more on judgment, information evaluation, planning, and leadership under uncertainty.

Data Points: U.S. manufacturing job losses: about 3.7 million - Loss of manufacturing employment between 1999 and 2007 cited as the scale of the China shock era decline. China shock period: 2001-2007 - Described as the main period when Chinese exports to the U.S. surged most dramatically. Follow-up study period: 2000-2019 - Used in the paper 'Places versus People' to examine longer-run labor-market adjustment. U.S. workforce with a four-year college degree: only 40% - Used to argue AI should help non-elite workers do more valuable work. Agriculture share of U.S. workers at start of 20th century: 38% - Historical comparison to show how automation transformed labor over time. Agriculture share of U.S. workers today: under 2% - Illustrates long-run labor substitution and structural change. U.S. healthcare spending paid by Medicare: 40% of U.S. healthcare - Used to show the scale of public involvement in healthcare and AI’s potential role there. Current manufacturing employment share: about 10% - Referenced as the baseline from which the U.S. should avoid further decline. Desired manufacturing employment floor: not falling from 10% to 5% - Otto warns against policies that accelerate decline in strategic manufacturing capacity.

Pivotal Quotes: "We are mounting our troops with maximum force to fight the last war." — David Otto: On Trump’s tariffs, arguing they target obsolete low-tech sectors rather than the real future competitive threat. "The right question is: you know, what did we do wrong back in the 2000s that this was so traumatic for our workforce?" — David Otto: Explaining that policy should focus on lessons from the China shock rather than nostalgia for lost industries. "AI will not do things on its own. It works for us, and we have to decide what we want it to do." — David Otto: On the need for deliberate design and governance of AI adoption.

Implications: Tariffs won’t rebuild the old industrial base; investment, skills, and allied industrial strategy matter more. AI could broaden opportunity or deepen inequality depending on whether policy steers it toward augmentation, not just labor-cutting.