The Turing Trap: The Promise & Peril of Human-Like Artificial Intelligence
Brynjolfsson draws a clean distinction between AI that substitutes (automation) and AI that augments (augmentation).
When AI imitates the human and replaces them, workers lose bargaining power and value concentrates.
When AI augments human capabilities, people keep the ability to capture value and new products and services appear.
The problem is that incentives push excessively toward automation — from technologists, executives and regulators alike.
The paper proposes rebalancing those incentives to avoid a concentration of economic and political power.
Central argument
Brynjolfsson argues that the AI field has fallen into a 'Turing Trap': an excessive focus on building systems that replicate human capabilities rather than ones that extend them. This distinction — automation versus augmentation — has concrete economic consequences: substitution erodes workers' bargaining power and concentrates value, while augmentation preserves human agency and generates new markets. The core thesis is not that automation is inherently bad, but that current incentive structures across technologists, executives, and regulators systematically bias investment toward substitution, and that deliberately rebalancing those incentives is both possible and necessary to avoid dangerous concentrations of economic and political power.
Critique
The automation/augmentation dichotomy, while analytically useful, risks being cleaner on paper than in practice — many real deployments sit somewhere in between, or start as augmentation and drift toward substitution as systems mature and costs fall. Brynjolfsson identifies misaligned incentives as the root cause but is more persuasive in diagnosing the problem than in specifying which policy levers would actually shift those incentives at scale without creating new distortions. A skeptical reader might also note that 'augmentation' can itself concentrate value if the humans being augmented are already highly skilled and well-compensated, leaving the distributional problem largely intact.
Why it matters for product
For a CPO, this framework reframes a decision that appears constantly in product and platform design: when you automate a user workflow entirely versus when you keep a human in the loop, you are not just making a UX or cost choice — you are making a structural choice about where value accrues and who retains leverage. Brynjolfsson's incentive argument is a direct warning about how roadmap prioritization, success metrics anchored to efficiency or headcount reduction, and vendor narratives can quietly bias a product organization toward substitution without anyone explicitly deciding that. The augmentation lens also opens a concrete product strategy question: what new services become possible when your users can do things they previously could not, rather than simply doing the same things with fewer people?