Andrew Ng: AI Will Not Cause Job Apocalypse, Software Engineers Still in Short Supply

TEXT START: Former Baidu Chief Scientist and founder of DeepLearning.AI, Andrew Ng, stated in an interview with The Economist that there is currently a shortage of software engineers, and there is also high demand for marketing, sales, and operations personnel who understand how to use AI.

The Dissection

This text is not a rebuttal of the Discontinuity Thesis. It is a snapshot of a transition phase repackaged as a systemic verdict. It shifts the unit of analysis from jobs to tasks, then treats surviving task fragments and current hiring shortages as proof that the job categories are permanently safe.

The punch-card analogy performs the same trick: it assumes AI is merely another productivity tool inside a human labor process, rather than a system that increasingly absorbs the process itself.

The Core Fallacy

It confuses present complementarity with permanent human indispensability. AI may automate only part of a job today because deployment, coordination, verification, and institutional adaptation lag behind capability. That does not establish that the remaining tasks require humans indefinitely.

Under P1, AI progressively dominates cognitive work. Under P2, institutions cannot preserve stable human-only economic domains at scale. Under P3, productivity gains eventually reduce the number of humans economically required. Current demand for AI-literate workers can therefore represent an implementation bottleneck—not durable labor scarcity.

Hidden Assumptions

• Current hiring demand will persist after AI deployment matures.
• Demand for exceptional engineers represents the security of ordinary engineers.
• Remaining judgment and coordination tasks cannot themselves be automated.
• New AI-related jobs will expand fast enough to offset displaced labor.
• Retraining can create enough economically necessary roles for the displaced majority.
• Wage labor will remain the mechanism that distributes purchasing power.
• A catastrophe only counts if unemployment reaches 50 percent or produces street protests.

That last assumption is especially dishonest. The system can die through wage compression, underemployment, labor-market exclusion, and concentration of productive ownership long before the headline threshold is reached.

Social Function

Primary classification: transition management, ideological anesthetic, and partial truth.

The partial truth is that task automation can initially increase demand for workers who deploy and supervise AI. The anesthetic is presenting that temporary demand as evidence that productive participation remains secure. The message directs workers toward retraining while leaving the ownership structure of AI capital untouched.

The Verdict

Ng’s claim defeats only a cartoon version of the job apocalypse: immediate 50 percent unemployment and mass street protests. It does not defeat the Discontinuity Thesis.

Software engineers being busy today is not proof against obsolescence. It is evidence of a lag phase in which AI makes a smaller number of workers more productive, concentrates capability, and creates temporary implementation demand. The relevant question is not whether jobs exist during deployment. It is whether human labor remains economically necessary after AI absorbs the cognitive production, verification, maintenance, and coordination around those jobs.

This text mistakes smoke from the burning labor circuit for proof that the circuit is healthy. It is a polished lullaby for a workforce being repriced.