Cheaper AI could create more jobs, not fewer: Apollo's chief economist Torsten Sløk- What he means

URL SCAN: Cheaper AI could create more jobs, not fewer: Apollo's chief economist Torsten Sløk- What he means
FIRST LINE: Fears that AI could trigger a wave of job losses may be overstated, according to Apollo Global Management chief economist Torsten Sløk.

TEXT ANALYSIS: "Zero Evidence of Job Losses Because of AI"

The Dissection

This article functions as a prestige signal embedded in the infrastructure of investor-class reassurance. It aggregates Sløk's Jevons paradox argument—the claim that AI-driven efficiency gains will expand markets and create net positive employment—and positions it against what the headline frames as "fears," deliberately rendering skeptics as anxious rather than analytical. The Reuters Altman quote serves as corroborating authority. The arXiv paper gets a single dismissive sentence as an afterthought, then immediately buried. The structural pattern is: leading optimists cited, dissenting data acknowledged but minimized, conclusion pre-sold.

The Core Fallacy

Sløk is measuring the wrong thing with the wrong time horizon and drawing the wrong conclusion. The argument that "zero evidence exists" in current ADP payroll data is the argument that a body isn't dead because rigor mortis hasn't set in yet. The Discontinuity Thesis does not predict immediate mass unemployment. It predicts the severing of the mass employment → wage → consumption circuit as the structural outcome of AI achieving durable cost and performance superiority across cognitive work domains. Current hiring of "AI implementation experts" and "data center buildout" is the transition phase, not the endpoint. It is the equivalent of noting that buggy whip manufacturers hired engineers to design better buggies in 1908 and concluding the horse had nothing to worry about.

The Jevons paradox is a real economic phenomenon. It is also historically bounded—it describes efficiency gains within a technological paradigm where human labor remains the marginal productivity driver. When the technology is the labor, Jevons paradox breaks down because there is no residual human input to scale. Sløk is applying a 19th-century蒸汽-era framework to a 21st-century cognitive automation process.

Hidden Assumptions

1. The transition is the destination. The article treats current AI-related hiring as evidence of net job creation without acknowledging that these roles are simultaneously being automated. AI implementation experts are themselves AI-replaceable within a defined and shrinking time horizon.
2. Official statistics capture structural displacement in real time. They do not. Labor market data lags structural reality by 2-5 years minimum, as the arXiv paper cited (and immediately dismissed) actually suggests. The paper finding employment challenges in AI-exposed occupations predating ChatGPT is the signal—the warning shot that the statistics are lagging a process already underway.
3. AI is currently creating more demand than it destroys across the full economy. Sløk points to data center construction, semiconductor demand, and specialist hiring. These are input industries to AI deployment, not evidence that AI deployment is net-positive across all sectors. The article never quantifies the ratio.
4. Technological revolutions of the past are structurally analogous. The industrial revolution replaced physical labor with machines but created demand for more physical labor at expanded scale. The cognitive automation revolution replaces the cognitive labor that generated the wages that created the demand. The analogy fails at the level of economic mechanism.

Social Function

Lullaby with a financial branding watermark. This article's primary function is to reassure investor confidence and consumer spending morale. Apollo Global Management is a major asset manager. Its chief economist publishing reassurance about AI and employment serves institutional interests in maintaining asset valuations premised on continued economic expansion. The article is not neutral journalism—it is uncritically amplifying a position that happens to align with the financial interests of the source's employer. The "growing debate" framing implies unresolved epistemic parity when the structural analysis is decidedly one-sided.

The Verdict

Sløk is diagnosing the patient in the waiting room and concluding the surgery was a success. The data he cites—ADP payrolls, current hiring rates, data center capex—are lag indicators of a transition, not evidence against a structural endpoint. The Discontinuity Thesis does not require mass unemployment to be visible in today's ADP data. It requires that the mechanism (AI achieving durable cost-performance superiority in cognitive work) is on a trajectory toward making human cognitive labor economically redundant at scale. That trajectory is not measured in quarterly payroll revisions. It is measured in the rate of capability improvement, the compression of deployment timelines, and the breadth of cognitive domains being automated. Sløk's "zero evidence" is an artifact of looking at the wrong data, with the wrong tools, on the wrong timescale.

The article performs a secondary function: intellectual cover for delay. Every reassurances piece buys another year of institutional inertia, regulatory inaction, and individual undersaving. That is not a coincidence.