AI was supposed to destroy jobs. Where's the carnage? - The Guardian

TEXT START: The prediction was stark: artificial intelligence advancements would wipe out jobs en masse.

THE DISSECTION

The article is not testing whether AI is displacing labor. It is testing whether displacement has already produced a synchronized spike in headline unemployment—a deliberately weak test.

Its own evidence describes the real process: hiring suppression, role consolidation, higher entry thresholds, smaller teams producing vastly more output, deteriorating job quality, and expanding contract work. The article’s actual finding is not that AI has failed. It is that AI has not yet converted productivity gains into mass visible unemployment.

THE CORE FALLACY

It confuses mechanical death with social death. Under the Discontinuity Thesis, the system breaks when AI destroys the necessity of mass human participation, not when every displaced worker is immediately recorded as unemployed.

Employers can absorb the shock through hiring freezes, attrition, role consolidation, productivity compression, and contractorization. A stable unemployment rate can coexist with a collapsing pipeline into durable careers.

The Stanford comparison is also a weak negative result. Exposed-worker unemployment not rising faster than least-exposed unemployment does not disprove delayed adoption, statistical lag, occupation-level variation, or displacement appearing first in hiring, wages, hours, job quality, and career ladders. The article itself supplies evidence for those channels.

HIDDEN ASSUMPTIONS

• Stable payroll employment means stable productive participation.
• A changed job remains economically necessary.
• New AI implementation, sales, and maintenance roles will scale enough to offset automation.
• Workers can upgrade skills faster than AI raises the performance threshold.
• Contract and freelance work is adaptation rather than liquidation of bargaining power and career formation.
• Institutional delays can reverse the process rather than merely postpone it.
• Current statistics can see a transition whose first effects occur in hiring decisions and job design.

SOCIAL FUNCTION

Partial truth functioning as transition management and ideological anesthetic. The reporting is valuable where it documents early displacement channels. Its anesthetic function is converting “not yet visible as mass unemployment” into “not much is happening.” Augmentation, turbulence, and skills matching provide respectable labels for fewer entry points, higher output demands, weaker career ladders, and more disposable labor.

THE VERDICT

The carnage is not absent. It is upstream of the unemployment headline.

The three-person analytics team reportedly producing the output of a 30-to-40-person team is not evidence that labor has been saved. It is evidence of labor compression. The missing jobs first appear as jobs never posted, junior workers never hired, teams never expanded, and careers downgraded into contracts.

The article mistakes the quiet phase of obsolescence for exoneration. Physical systems, organizational inertia, political resistance, and bad measurement are lag defenses. They can delay the break; they cannot refute the mechanism. If P1 and P2 hold, P3 will not arrive as one cinematic layoff event. It will arrive as the slow removal of economic necessity from the majority while employment statistics continue reporting that many people technically have jobs.