The case for AI as an engine of job growth, not a layoff machine - AOL.com

URL SCAN: The case for AI as an engine of job growth, not a layoff machine - AOL.com
FIRST LINE: A growing chorus of economists led by Torsten Slok is making the case for AI creating job growth.

THE DISSECTION

This article is a prestige-hedging exercise dressed as economic analysis. It exists to launder a specific ideological product: the comfort that AI will be "additive" like past technological transitions, so the reader doesn't panic, sell, or demand regulation. The radiologist example is the centerpiece artifact—hand-picked, temporally truncated, and circular. The China shock parallel is the secondary drug, used to anchor the reader in a historical comparison that breaks down the moment you examine its structural preconditions.

THE CORE FALLACY

Slok's Jevons Paradox argument commits a category error: it assumes that because lower costs can expand demand in a market, this mechanism operates uniformly across all labor sectors under all technological conditions. It doesn't.

Jevons works when the bottleneck is supply capacity, not when the bottleneck is human cognitive participation in the value chain itself. Radiologists reading more scans because AI handles triage is a bounded domain. Software engineers being replaced by AI that writes, tests, deploys, and maintains its own code is not the same dynamic. One is task augmentation within a human-anchored workflow. The other is structural displacement of the worker from the value-creating circuit entirely.

The article concedes this—"Slok compared it to 'very regional-specific' disruptions" for software—but then immediately walks it back by insisting new AI-adjacent jobs will offset. What it never addresses: the productivity-to-employment ratio in cognitive work is not 1:1. It has never been 1:1. And it is trending sharply negative.

HIDDEN ASSUMPTIONS

1. Historical continuity of labor market response. The China shock disrupted manufacturing, but that occurred within a labor market where displaced workers could move into services, construction, and logistics—sectors where human physical coordination remained the productive bottleneck. AI targets the cognitive coordination bottleneck itself, which means the secondary absorption pathways are structurally foreclosed.

2. Net job creation is the relevant metric. Even if AI-adjacent jobs are created, the Discontinuity Thesis does not require all jobs to be destroyed. It requires that the majority lose access to economically necessary productive participation. You can have net job growth while the distribution of those jobs makes them inaccessible to the displaced population in terms of skills, geography, or capital requirements.

3. The radiology example is current, not historical. Radiologists making $500k "today" is doing enormous work to conceal the direction of travel. The article was written at a moment when radiology AI is still in adoption lag. The displacement wave for diagnostic radiology is demonstrably accelerating in 2024-2025. The "more radiologists needed" argument is the 2019 talking point. It is now the thing people cite because they're afraid to cite the thing that's actually happening.

4. Corporate AI scapegoating is a separate phenomenon from actual AI displacement. The article suggests companies may be "using AI as a scapegoat" for mismanagement. This is almost certainly true in some cases. But the logical structure of this argument is: "some companies are lying about AI causing layoffs, therefore AI isn't causing layoffs." The existence of motivated misreporting does not establish the counterfactual.

SOCIAL FUNCTION

Copium. Specifically, institutional copium targeted at middle-class knowledge workers who are the primary readership of AOL and Business Insider. The article tells them: your job is safe, the economists say so, and even when companies say they're cutting for AI, they might be lying to you. It preserves consumption patterns and political quiescence by managing anxiety rather than encoding accurate threat assessment.

THE VERDICT

Torsten Slok is selling a thesis that was marginally defensible in 2019 and is now empirically underpowered. The China shock analogy is a category error on the order of comparing a fire to a controlled burn. The radiology example is a lagging indicator being deployed as a leading indicator. The article's own evidence—software layoffs at Salesforce, Atlassian, IBM—undermines its headline thesis in the third paragraph, and the rest of the piece performs elaborate narrative repair to avoid acknowledging what it just reported.

Structural reality under DT logic: AI displacement in cognitive work is not "very regional-specific" in the sense of manufacturing's geographic concentration. It is sector-wide, function-wide, and structurally permanent once the AI achieves task-completion parity at below-human marginal cost. The Jevons expansion effect requires new human jobs to fill expanded demand. AI does not require humans to fill expanded demand. This is the mechanical difference Slok's model does not incorporate.