“Comment on Barentt (2025): Growth effects of AI could hit a bottleneck even if local elasticities are high” by Dan Carey

Introduction This post is a response to a recent article by Matthew Barett on the Epoch AI website, posted as part of the Gradient Updates newsletter. If you can, I recommend skimming Barnett's piece before reading further. I will refer to it throughout as Barnett (2025).[1] Brief Summary of Barnett (2025) Barnett (2025) uses an aggregate production function to estimate how non-universal task automation might affect output (GDP). To identify which tasks are likely to be automated, he relies on a GPT-4o-based classification that determines which tasks can be performed remotely (finding about 34% fit that description). He assumes that “remote” tasks largely overlap with “automatable” ones. To estimate how easily automatable and non-automatable tasks can substitute for each other, he draws on the shift to remote work during the COVID era. The idea is that tasks once performed in person were abruptly carried out online, letting us [...] ---Outline:(00:09) Introduction(00:31) Brief Summary of Barnett (2025)(01:41) Brief Summary of this Critique(03:30) Theory(14:35) Empirics(14:52) Endogeneity of the optimistic case(15:53) The pessimistic case is not a good lower-bound(17:14) Wage share(18:04) ConclusionThe original text contained 6 footnotes which were omitted from this narration. --- First published: March 18th, 2025 Source: https://forum.effectivealtruism.org/posts/DMxsAGtyhhnzXoe27/comment-on-barentt-2025-growth-effects-of-ai-could-hit-a --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.