Generative AI is an added layer, not a rupture: it integrates machine co‑creation into existing information infrastructures, amplifying complementarities, governance challenges and platform advantages rather than outright replacing prior systems.

Main Finding

Generative AI is best understood not as a sudden, revolutionary rupture but as the latest interim layer in a long-running evolution of information infrastructures. It adds machine co-creation to an already layered information landscape (pre‑digital → centralized computing → personal computing → web → social media), increasing complexity and interaction among layers rather than fully displacing prior systems.

Key Points

• Historical continuity: Information systems have evolved in successive layers; each layer builds on and coexists with predecessors while introducing new capabilities and frictions.
• Layer characteristics: Each layer can be described by its infrastructure (who controls storage/processing), information flows (one‑to‑many, many‑to‑many, algorithmic mediation), and human roles (producer, curator, consumer, moderator).
• Novelty of generative AI: It positions machines as active co‑creators within hybrid human‑machine systems, enabling automated content generation, personalized synthesis, and new forms of interaction.
• Coexistence and complexity: Generative AI augments existing platforms and practices, producing interactions across layers (e.g., AI‑generated content distributed via social media) and new governance, attribution, and trust challenges.
• Normative point: To realize benefits while limiting harms, policy and design should aim to augment human agency and manage sociotechnical risks (misinformation, concentration, erosion of labor tasks).

Data & Methods

• Nature of study: Conceptual and historical synthesis rather than empirical causal analysis. The paper constructs a comparative framework by tracing technological, organizational, and social features across successive information eras.
• Methods used: Historical comparison, taxonomy of “layers” (infrastructure, flows, roles), and qualitative argumentation about complementarities and frictions introduced by each layer.
• Limitations: No primary quantitative data or causal identification. Empirical validation is needed to measure the magnitude and heterogeneity of generative AI’s economic effects.

Implications for AI Economics

• Complementarity vs substitution: Economists should expect mixed effects—generative AI will substitute for some routine cognitive tasks while complementing skilled workers and creative tasks, shifting the skill composition of labor demand.
• Productivity measurement: Standard metrics (hours worked, output) may understate value from AI‑augmented work (quality, speed, variety). New measurement approaches are needed for AI‑generated outputs, attribution, and consumer surplus.
• Market structure & platforms: Generative AI capabilities can increase returns to data and model ownership, reinforcing platform advantages and potentially increasing market concentration; regulatory and competition considerations matter.
• Information goods & attention markets: Faster, cheaper content production changes supply dynamics, attention scarcity intensifies, and content moderation/externalities become central economic frictions.
• Distributional consequences: Potential for widening inequality (returns to capital/data owners, skill‑biased gains) unless complemented by policies (retraining, access to tools, redistribution).
• Policy and institutional design: Focus on augmenting human agency—standards for provenance/attribution, liability rules, taxation and intellectual property adjustments, support for worker transitions, and competition policy to avoid lock‑in.
• Research agenda for economists:
  • Measurement: develop firm‑ and task‑level indicators of AI adoption and output quality.
  • Identification: exploit natural experiments, staggered rollouts, and randomized interventions to estimate causal impacts on productivity, employment, and prices.
  • Market models: study dynamic competition with large fixed costs (models, data), platform ecosystems, and multi‑sided markets.
  • Distributional and welfare analysis: quantify consumer surplus, labor income changes, and externalities (misinformation, privacy).
  • Policy evaluation: analyze regulation, subsidies for complementary investments (training, interfaces), and tax/transfer designs.

Overall, treating generative AI as an additional, interacting layer helps economists focus on complementarities, measurement challenges, market dynamics, and policy levers needed to steer technological change toward broadly shared gains.