Gary Marcus Questions Generative AI’s Human Payoff

An October 24 essay revisits the balance of benefits and damage from widely deployed generative systems.

Gary Marcus’s October 24 essay asks an unfashionably basic question: is generative AI improving human life enough to justify its damage? He points to nonconsensual sexual imagery, fraud, misleading information, surveillance and pressure on creative work. He also acknowledges coding benefits and employment effects that have been less severe than some expected. [s1]

Its importance is that it refuses to count technical capability as a benefit automatically, before examining what that capability actually does to people.

The public debate too often treats harm as a temporary engineering expense on the way to an extraordinary future. That framing asks people facing exploitation now to accept a speculative repayment later. A more serious moral accounting would name the people receiving the benefits and those paying the price, then ask whether either group had a meaningful choice.

There is no single exchange rate between faster drafting and a damaged reputation, or between cheaper content and the erosion of a creative livelihood. Those questions involve dignity, responsibility and the ability to seek redress. They cannot be settled by counting how many users adopted a product.

The constructive response is to separate uses that deliver demonstrable help from uses that depend on exploitation or leave victims without recourse. A society can support carefully bounded assistance while refusing to normalise abuse as the cost of innovation. That preserves a hopeful possibility for AI: earning a place through service to human purposes, instead of demanding acceptance because a business model requires scale.