teleo-codex/inbox/archive/2026-02-27-noahopinion-roundup-78-roboliberalism.md

type	title	author	date	url	domain	processed_by	status	notes	claims_extracted
source	Roundup #78: Roboliberalism	Noah Smith (Noahopinion)	2026-02-27	https://www.noahpinion.blog/p/roundup-78-roboliberalism	internet-finance	rio	processed	Section 1 (AI productivity) is relevant. Sections 2-4 (housing, global poverty, American wealth) are outside Rio's domain.	Current productivity statistics cannot distinguish AI impact from noise because measurement resolution is too low and adoption too early for macro attribution Early AI adoption increases firm productivity without reducing employment suggesting capital deepening not labor replacement as the dominant mechanism

Roundup #78: Roboliberalism

Noah Smith's newsletter roundup, Feb 27, 2026. Only Section 1 on AI and productivity is relevant to internet finance extraction.

Section 1: Is AI boosting productivity?

Brynjolfsson's 2.7% claim

Erik Brynjolfsson claims US productivity growth hit 2.7% in 2025 — nearly double the 1.4% average. His evidence: the BLS revised payrolls down by 403K while GDP grew 3.7%. If output grew while labor input shrank, that's a productivity surge. He attributes this to AI adoption.

Counter-evidence #1: Gimbel

• The data is noisy — 403K revision sounds large but is within normal revision ranges
• GDP itself gets revised, often substantially
• Immigration policy changes (deportations, reduced legal immigration) confound the labor supply picture
• Capital investment broadly (not AI specifically) could explain productivity gains
• "We simply don't have the statistical resolution to attribute productivity changes to AI at this point"

Counter-evidence #2: Imas (productivity J-curve)

• Technology transitions follow a J-curve: initial dip as workers learn new tools, then acceleration
• Micro-level AI productivity gains exist but aren't yet large enough to move macro statistics
• The absence of macro evidence doesn't disprove the thesis — it may just be too early
• Historical parallel: computers didn't show up in productivity statistics until the late 1990s, decades after adoption began (Solow paradox)

Counter-evidence #3: Aldasoro et al (BIS/EU study)

• AI-adopting firms show ~4% productivity increase
• But NO evidence of employment reduction at those firms
• Interpretation: AI is capital deepening (making existing workers more productive) not labor replacement
• This directly challenges the displacement-as-default thesis
• If AI makes workers more productive without replacing them, the macro crisis scenario requires a different mechanism

Counter-evidence #4: Yotzov survey (6000 executives)

• Executives report small current impact from AI
• Expect 1.4% productivity boost and 0.7% employment cut over coming years
• The expected employment effect is roughly half the productivity effect
• This is modest compared to both the catastrophist and utopian narratives

Noah's synthesis

"We don't really know how technology affects productivity, growth, employment, etc. until we try it and see." The honest position is radical uncertainty — neither the doom narrative nor the boom narrative has sufficient empirical support. Current data is too noisy, adoption too early, and measurement too crude to distinguish between the competing macro scenarios.

Sections 2-4 (not extracted)

• Section 2: Yuppie Fishtank Theory (housing/urban economics)
• Section 3: Global poverty trends
• Section 4: American wealth distribution These are outside internet-finance domain scope.