## A New Industrial Revolution?

## Source details

**Canonical URL:** [A New Industrial Revolution?](https://www.imf.org/en/publications/fandd/issues/2025/12/a-new-industrial-revolution-niall-kishtainy)

## Other formats

- [Markdown version](/en/publications/fandd/issues/2025/12/a-new-industrial-revolution-niall-kishtainy/index.md)
- [Structured JSON version](/en/publications/fandd/issues/2025/12/a-new-industrial-revolution-niall-kishtainy/index.json)
- [Bundle manifest](/en/publications/fandd/issues/2025/12/a-new-industrial-revolution-niall-kishtainy/bundle-manifest.json)

## Bibliographic details
- Authors: NIALL KISHTAINY
- Published: December 3, 2025

---

### Overview
- Author: NIALL KISHTAINY
- Publication: F&D Magazine, December 2025
- Thesis: AI may constitute a distinct general purpose technology potentially as transformative as steam, electricity, and ICT, but history suggests its full economic impact will take time and will depend on organization, distributional effects, and policy choices.

### General purpose technology and timing
- Definition: General purpose technology has many uses and increases productivity throughout the economy; it often involves a new “method of invention.”
- Historical parallels:
  - First Industrial Revolution: steam power (late 18th century → Britain).
  - Second Industrial Revolution: electricity (beginning late 19th century).
  - Third Industrial Revolution: ICT (beginning late 20th century).
- AI assessment:
  - AI is likely a distinct general purpose technology and a new method of invention.
  - Historical lesson: full effects of such technologies require broad economic reorganization and take time.
  - Empirical note: Nicholas Crafts found steam’s gains came only after 1830 because early sectors using steam were a small fraction of the economy.

### Productivity paradox and pace of effect
- Solow paradox: "you can see the computer age everywhere but in the productivity statistics."
- Experience with ICT:
  - ICT’s productivity gains were historically unprecedented in speed and magnitude compared with early steam.
  - AI’s practical economic application is still at an early stage; AI’s contribution to productivity so far has been modest and has already been called a “productivity paradox.”
  - If the ICT experience is a guide, AI’s productivity impact may be felt faster than earlier general purpose technologies, though not necessarily yielding the most optimistic growth projections.
- Short-term vs long-term drivers:
  - Over centuries, technology has driven growth and living standards (supply-side improvements).
  - In the short term, factors such as weak demand (noted after the global financial crisis of the early 2000s) and the exhaustion of “low-hanging fruit” innovations (Robert Gordon’s argument) influence observed growth rates.

### Distributional effects: displacement, reinstatement, and inequality
- Displacement effect:
  - New technologies automate tasks, leading to shedding of labor and reduced labor share of national income (Acemoglu and Restrepo).
  - Historical example: mechanization of weaving devastated hand-loom weavers in mid-19th century Manchester; many survived on collapsing wages and 18-hour workdays.
- Reinstatement effect:
  - Technologies can create new tasks where humans have comparative advantage, generating jobs (e.g., engineers, telephone operators, machine technicians, software designers).
  - The reinstatement effect became stronger in the second half of the 19th century and during the 20th century, driving up wages and living standards.
- Net patterns:
  - Early Industrial Revolution: output per worker rose while real wages stagnated; wages rose in line with productivity only after the middle of the 19th century (Robert Allen).
  - Late 20th century onward: real wages in many leading economies have been flat despite productivity gains.
  - Recent AI and many ICT innovations have been aimed at automation rather than creating new tasks, exacerbating stagnant labor demand, slow wage growth, and rising inequality.
- Policy recommendation from cited scholars:
  - Pursue labor-reinstating AI (examples: education and health) where AI tools could require more human labor by enabling individually tailored learning and treatment programs.

### Singularity question and limits of substitution
- Machine singularity concept: machines could improve and invent themselves, potentially eliminating labor reinstatement.
- Empirical skepticism:
  - William Nordhaus’s empirical tests suggest most conditions for economic singularity are far from being met.
  - Much of the economy is physical, not informational; full takeover would require machines to perform many physical, routine, and caregiving tasks (examples: poach eggs, cut hair, soothe crying toddlers).
- Implication: even dramatic AI advances would not automatically make economic comparisons with earlier eras useless.

### Policy implications and social choice
- Market failure and social choice:
  - Innovation involves significant market failures.
  - Currently, choices about AI’s path are largely being left to corporations with limited concern for broader economic impacts.
- Role for government and regulators:
  - Compared with the early 19th century, we now possess policy tools to influence technological outcomes.
  - Countries have both motive and means to guide technological development to ensure broadly shared economic benefits—contingent on political will.

*Source: NIALL KISHTAINY, "A New Industrial Revolution?", F&D Magazine, December 2025.*

---

## Content in this bundle

- **A New Industrial Revolution?**
  - [A New Industrial Revolution? (Markdown version)](/-/media/files/publications/fandd/article/2025/12/kishtainy.pdf.md){rel="alternate" type="text/markdown"}
  - [A New Industrial Revolution? (PDF)](/-/media/files/publications/fandd/article/2025/12/kishtainy.pdf){rel="external" type="application/pdf"}

---

_Source: https://www.imf.org/en/publications/fandd/issues/2025/12/a-new-industrial-revolution-niall-kishtainy_