## AI’s Promise for the Global Economy

## Source details

**Canonical URL:** [AI’s Promise for the Global Economy](https://www.imf.org/en/publications/fandd/issues/2024/09/ais-promise-for-the-global-economy-michael-spence)

## Other formats

- [Markdown version](/en/publications/fandd/issues/2024/09/ais-promise-for-the-global-economy-michael-spence/index.md)
- [Structured JSON version](/en/publications/fandd/issues/2024/09/ais-promise-for-the-global-economy-michael-spence/index.json)
- [Bundle manifest](/en/publications/fandd/issues/2024/09/ais-promise-for-the-global-economy-michael-spence/bundle-manifest.json)

## Bibliographic details
- Authors: MICHAEL SPENCE
- Published: September 3, 2024

---

### Core thesis
- If properly used, AI could significantly accelerate economic growth and help productivity growth rebound.
- Meaningful impacts in labor productivity may begin to appear "by the end of this decade" (author’s best guess).
- Roy Amara’s law: likelihood of overestimating short-run impacts and underestimating long-run impacts of technological transformation.

### Major headwinds and shocks
- Colliding shocks reducing supply elasticity and raising costs:
  - War, pandemic, climate change, geopolitical tensions, resurgent nationalism, and national-security–focused economic policy.
  - Rapid postpandemic fragmentation of global supply networks driven by diversification and resilience priorities, and policy initiatives to bring supply chains home or to friendly countries.
- Examples and structural consequences:
  - India now produces 15 percent of iPhones.
  - Only South Korea and Taiwan Province of China make (as opposed to design) the most advanced semiconductors.
  - Tradeoffs: cannot maximize resilience and minimize costs simultaneously; structural shift has contributed to inflationary pressures.

### Secular trends and productivity dynamics
- Key secular forces reducing supply elasticity:
  - Declining productivity, especially in advanced economies.
  - Aging populations in economies that account for more than 75 percent of global output.
  - Rising sovereign debt following the pandemic; global sovereign debt now exceeds global gross domestic product.
- Exact sovereign debt ratios cited:
  - United States: 120 percent.
  - Europe: 88.6 percent (with Greece, Italy, Spain, France, Belgium, and Portugal above this average; Greece and Italy "by a lot").
- US productivity statistics (exact figures preserved):
  - US productivity growth averaged 1.68 percent from 1998 to 2007.
  - Productivity growth slowed to 0.38 percent from 2010 to 2019.
  - Tradable goods and services sectors: fell from 4.27 percent to 1.23 percent.
  - Nontradable services sectors: declined from 0.73 percent to effectively zero.
- Measured productivity edged up during the pandemic due to partial shuttering of less productive industries and shift to remote work in higher-productivity sectors.
- Structural outcome: relatively rapid shift from demand-constrained to supply-constrained growth — subdued growth, enduring inflation, elevated real interest rates, and likely higher borrowing costs than the decade following the global financial crisis.

### Technological revolutions and AI’s potential
- Three revolutionary transformations:
  - Multidecade digital transformation accelerated by breakthroughs in AI.
  - Revolution in biomedical and life sciences.
  - Technologies underpinning the transition to sustainable energy.
- Characteristics of generative AI:
  - First AI with humanlike capacity to operate in multiple domains and detect/switch domains based on conversational prompts.
  - Capable of tasks such as discussing inflation, writing computer code, and doing some mathematics (noted as work in progress).
  - Better framed as machine-human collaboration ("augmentation") rather than full automation.
- General-purpose nature:
  - AI as a general-purpose technology with applications across the economy; only general-purpose technologies can produce economy-wide productivity surges.
- Potential sectoral impacts:
  - High-impact sectors already investing heavily (technology, finance).
  - Need diffusion to large employment sectors that tend to lag: government, health care, construction, hospitality.

### Challenges to achieving potential
- Key barriers:
  - Talent, computing power, and rapidly expanding electricity demand for training increasingly powerful generative AI models.
  - Regulatory need to prevent misuse of technology and data; risk-mitigation regulatory agenda is underway globally.
  - Automation bias (the "Turing Trap"): tendency to view AI as full automation and replacement for humans.
  - Concentration of powerful training systems in private-sector cloud computing (mostly in the US and China) and competition for talent disadvantaging science and academia.
- Data availability:
  - Internet provides ample training data; personalized and sensitive data are not required to train large language models.
  - Specialized applications (e.g., AlphaFold) require domain-specific data and expert input.
- Geographic and policy risks:
  - Europe risks falling behind the United States and China for three reasons:
    - Relative underfunding of basic research.
    - Lag in computing power to support research.
    - Failure to fully leverage the large scale of the European economy (fragmented capital markets and fragmented regulation).
  - China characterized as an AI powerhouse.
  - India likely a growing force given digital roots, large internal market, and engineering human capital.
  - Most other emerging market economies will be largely consumers of advanced AI technology in the near term.

### Policy recommendations and priorities
- Rebalance policy emphasis:
  - Strengthen policies for accessibility, diffusion, and skills acquisition alongside risk-mitigation and misuse prevention.
  - Avoid government "picking winners"; effective competition policy should be part of the portfolio.
  - Focus on diffusion to lagging sectors and support for small and medium enterprises.
  - Prioritize retraining and new skills acquisition as jobs change with AI collaborators.
- Democratize infrastructure and research:
  - Expand computing infrastructure to a broad community of researchers and innovators to balance academic and private innovation and support widespread diffusion.
- Anticipated macroeconomic benefits with supportive policy:
  - With policy support to accelerate diffusion across the entire economy, AI could significantly accelerate economic growth and help productivity rebound.
  - If AI relaxes supply-side constraints, it could indirectly lower real interest rates and the cost of capital over time, aiding the energy transition and supporting aging populations.

### Key projections and scenarios
- Timing:
  - Author’s best guess: meaningful impacts in labor productivity may begin "by the end of this decade."
- Diffusion scenarios:
  - If AI adoption remains concentrated in tech-intensive sectors, economy-wide gains are unlikely to be fully realized.
  - If AI is broadly accessible and diffused to lagging sectors with supportive policy, AI could produce a major sustained surge in productivity over time.

*AI’s Promise for the Global Economy — Michael Spence, F&D Magazine, September 2024.*

---

## Content in this bundle

- **Staff Paper**
  - [Staff Paper (Markdown version)](/-/media/files/publications/fandd/article/2024/09/spence.pdf.md){rel="alternate" type="text/markdown"}
  - [Staff Paper (PDF)](/-/media/files/publications/fandd/article/2024/09/spence.pdf){rel="external" type="application/pdf"}

---

_Source: https://www.imf.org/en/publications/fandd/issues/2024/09/ais-promise-for-the-global-economy-michael-spence_