AI job losses or new opportunities? Infosys cofounder Kris Gopalakrishnan weighs in

TEXT START: Gopalakrishnan made the point while responding to a viral video featuring AI pioneer Geoffrey Hinton, who revisited his earlier prediction that AI would largely replace radiologists.

The Dissection

The text converts one radiology rebound effect into a general defense of employment. Its structure is simple: AI makes scans cheaper and faster; cheaper services generate more demand; therefore automation may create more jobs than it destroys.

That is not an economic demonstration. It is reassurance built around a vivid anecdote. The article shifts attention from labor displaced per task to total service demand, then treats the existence of residual human responsibilities as evidence that the profession remains secure.

The Core Fallacy

The central error is confusing increased output with preserved human employment.

Radiology may experience demand expansion because lower costs produce more scans. That can delay displacement or create temporary demand for complementary workers. It does not refute the Discontinuity Thesis. If AI eventually reads nearly all scans, the remaining human work can be narrower, cheaper, and performed by fewer people—even while total scans increase.

The article mistakes a rebound effect for a reversal of automation. More economic activity does not guarantee that the human share of that activity remains large enough to sustain mass employment, wages, or bargaining power.

Hidden Assumptions

• Every sector will generate enough new demand to absorb displaced workers.
• New demand will translate into human jobs rather than mostly more AI-mediated output.
• Human duties such as patient communication and difficult-case consultation cannot themselves be automated, compressed, or concentrated among fewer specialists.
• The radiology pattern generalizes across the economy.
• “Jobs created” are equivalent to jobs lost in pay, status, stability, and accessibility.
• Aggregate demand will survive the erosion of the wage-to-consumption circuit.
• AI adoption will remain task-level rather than progressing toward broad cognitive labor substitution.

The quote “AI is creating more jobs than we lose to AI” is presented as a conclusion without figures, time horizon, sectoral accounting, or evidence about job quality. It is an assertion wearing the costume of a trend.

Social Function

Classification: partial truth, copium, and transition management.

The partial truth is real: automation can lower prices, expand consumption, and create complementary roles. The copium begins when that local mechanism is used to imply systemic labor-market safety. The text gives institutions and workers a comforting story in which displacement is merely transformation and every lost task is compensated by a new opportunity.

Its ideological function is to preserve faith in the existing employment order while the productive core is being automated. It does not ask who owns the AI, who captures the productivity gains, or whether the new opportunities are numerous and valuable enough to replace the destroyed wage base.

The Verdict

This is not a refutation of AI-driven labor obsolescence. It is a polished anecdote about demand expansion being mistaken for employment preservation.

Radiology demonstrates that automation can enlarge a market before it eliminates the labor bottleneck. Under the Discontinuity Thesis, that is lag—not survival. The decisive question is not whether humans remain somewhere inside the workflow. It is whether most humans remain economically necessary. This text provides no evidence that they will.