Is AI Really Taking Our Jobs? — With Noah Smith

Episode Summary

Executive Summary: Noah Smith argues that AI’s current productivity impact is still too small to detect clearly, and that major gains may require companies to reorganize work rather than simply replace humans with chatbots. He also contends immigration restrictions and tariffs are economically negative, while rising political grievance is driven more by social media and elite polarization than by worsening material conditions.

Main Topics: AI and productivity: early evidence is limited (Priority: 5/5): Smith says generative AI is too new for productivity statistics to show a clear effect, and that broad productivity growth since the pandemic is likely driven by other factors. He expects meaningful gains only when firms redesign workflows around AI, similar to how electrification transformed factories. Why AI hasn’t produced a business transformation yet (Priority: 5/5): He argues most companies are using AI as a human replacement tool, which is often the wrong mental model. The real gains, if they come, will require organizational changes, new task division, and complementary use of human labor and AI. Entry-level hiring and the AI labor-market narrative (Priority: 4/5): Smith says there is no solid evidence of an AI-driven white-collar job apocalypse. Slower entry-level hiring could reflect macro conditions, post-2021 overhiring, and cohort/major crowding rather than AI. Inequality, wages, and the hollowed-out middle class (Priority: 4/5): He rejects the idea that the middle class has been broadly hollowed out by technology. In his view, wage inequality has been falling since roughly 2013, with lower-end wages rising faster than upper-end wages, even though asset inequality remains large. Politics, grievance, and social media (Priority: 4/5): Smith attributes rising political unrest less to worsening living standards and more to social media, which concentrates conflict, amplifies outrage, and prevents people from socially sorting into more compatible communities. Immigration and tariffs as growth constraints (Priority: 5/5): He argues broad immigration crackdowns reduce disruption-reduction benefits from labor supply and demand, while skilled immigration is strongly positive for innovation. Tariffs, meanwhile, raise uncertainty, create bottlenecks, and slowly erode economic strength. US-China rivalry and strategic competition (Priority: 4/5): Smith says US-China competition is real and potentially dangerous, especially around Taiwan and military power. He argues the US should strengthen alliances and deterrence while also staying competitive in AI and other military-enabling technologies.

Key Arguments: AI’s economic impact is not yet visible in aggregate productivity data because mass business adoption is still limited and LLMs are only a couple of years old. The biggest productivity gains from new technology often come not from direct replacement, but from reorganizing production around the technology’s strengths, as with electricity in factories. Predictions that AI will eliminate large shares of jobs are overstated because they ignore hidden tasks, task recombination, and the creation of new work. Lower- and middle-income wages have risen in recent years, so current economic pain narratives do not match broad wage data as well as they did in the 1970s-80s. Asset inequality is real, but it is driven more by financial and market dynamics than by a simple productivity story. Social media intensifies conflict by forcing unlike-minded groups into constant contact and rewarding outrage. Immigration is economically positive overall because immigrants both supply labor and create demand; restricting skilled immigration is especially harmful because it reduces innovation and clustering effects. Tariffs create uncertainty and sector-specific harm even if markets assume Trump will back down; they can slowly weaken growth rather than cause an immediate crash. The US must maintain deterrence against China through alliances, military capability, and technological leadership, especially in AI and autonomy.

Data Points: AI age: 2 years - Smith says generative AI/LLMs have only existed for a couple of years, making productivity effects hard to measure. Productivity growth since the pandemic: Solid, not spectacular - He says US productivity has grown decently overall, though not clearly due to AI. Factory productivity gain from electrification: About 5x - Used as an analogy for how reorganizing production around a technology can unlock large gains. Company AI deployment with no earnings impact: Nearly 8 in 10 companies - Cited from a McKinsey report describing the 'generative AI paradox.' Companies employing AI to produce goods and services: Less than 10% - Smith cites unofficial measures suggesting limited real business adoption. Relative unemployment of young graduates: Started rising in 2009 - He cites economist reporting that this trend predates generative AI by many years. Actual unemployment rate of young graduates: Around 6% - Cited as elevated relative to some periods but not catastrophic. Occupations thought vulnerable to AI: Still rising steadily since about 2000 - Back office support, financial operations, and sales employment were not showing an AI disruption in the cited data. Wage inequality trend: Falling for over 10 years - He says lower wages have risen faster than higher wages since roughly 2013. Inequality run-up period: 1980s, 1990s, 2000s - He acknowledges wage inequality rose significantly during these decades. Trump tariff pause expiration: July 9 - Referenced as the deadline for the global tariff pause. Tax and spending bill date: July 4 - Mentioned alongside the tariff deadline in the news context. China nuclear weapons buildup: 100 nuclear weapons a year - Used to illustrate the scale of the strategic challenge from China.

Pivotal Quotes: "we don't know what the hell's going on" — Noah Smith: On whether AI is causing slower entry-level hiring and broader labor-market shifts. "I think that those dire predictions of what jobs that AI is going to destroy ... are bunk" — Noah Smith: On models that estimate large shares of jobs are vulnerable to AI by summing automatable tasks. "everyone was thrown in one little room with everyone else" — Noah Smith: Describing how social media intensified political and cultural conflict by collapsing social/geographic separation.

Implications: Listeners should expect AI gains to come from workflow redesign, not instant job replacement. Immigration, tariffs, and social-media-driven polarization may matter more for near-term economic and political outcomes than AI hype. The US-China race will hinge on alliances, deterrence, and applied AI, not just model quality.