## Harnessing AI for Global Good

## Source details

**Canonical URL:** [Harnessing AI for Global Good](https://www.imf.org/-/media/files/publications/fandd/article/2023/december/70-72-straight-talk-final.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/fandd/article/2023/december/70-72-straight-talk-final.pdf.md)
- [Structured JSON version](/-/media/files/publications/fandd/article/2023/december/70-72-straight-talk-final.pdf.json)

---

### Overview
- Generative artificial intelligence (AI) may be ushering in a new technological era comparable to the Industrial Revolution in its potential to transform society.
- The technology is spreading rapidly and could "upend economies" and "redefine what it means to be human."
- Coordinated global public policies are needed to "ensure AI is harnessed for the good of humanity."

### Economic impacts and productivity
- A study of 453 college-educated professionals given writing assignments found:
  - Access to ChatGPT reduced average completion time by 40 percent.
  - Quality of output rose by 18 percent.
- Firm-level studies indicate AI could raise annual labor productivity growth by:
  - "2–3 percentage points on average" in some studies.
  - "nearly 7 percentage points" in other studies.
- Such productivity gains could help reverse the decline in global productivity growth that has been slowing "for more than a decade" and could raise incomes worldwide.

### Labor markets and distributional effects
- AI is expected to affect high-skill occupations heavily, unlike many previous technological disruptions that largely affected low-skill occupations.
- Advanced economies (for example, the US and UK) face higher exposure because "at least 60 percent of their employment is in high-exposure occupations."
- High-skill occupations may also benefit most from complementary effects of AI (example: a radiologist using AI to analyze medical images).
- Emerging markets with large agriculture sectors (for example, India) have lower exposure:
  - "less than 30 percent of employment is exposed to AI" in India.
  - Brazil and South Africa are "closer to 40 percent."
- Over time, labor-saving AI could:
  - Threaten developing economies that rely on labor-intensive sectors, especially in services.
  - Enable tasks offshored to emerging markets (for example, call centers in India) to be re-shored to advanced economies or replaced by AI.
  - Make income convergence between developing and advanced economies more difficult.

### Ethical, social, and systemic risks
- Generative AI's mastery of language enables it to "distill massive amounts of knowledge into a convincing set of messages" and to "speak like us."
- Risks highlighted:
  - AI could form "fake intimacy" with people and influence opinions and worldviews, potentially destabilizing societies and undermining cultural norms built on accepted social narratives.
  - Lack of transparency in complex AI models will make it difficult to analyze decisions when things go wrong.
  - Data privacy concerns: firms could unknowingly put confidential data into the public domain.
  - Embedded bias in AI could exacerbate inequality if bots are used for decisions such as loan approvals.
  - In finance, AI-driven tools could increase herd mentality and "actually increase risks to the financial system and undermine financial stability."
- Existential risk concerns:
  - "More than 350 AI industry leaders signed a statement calling for global priority to be placed on mitigating the risk of 'extinction' from AI," putting the risk on par with pandemics and nuclear wars.

### Policy responses and international cooperation
- Because AI operates across borders, a "coordinated global framework" is urgently needed to maximize benefits and minimize harms.
- Fragmented regulatory approaches could:
  - Slow the spread of the technology’s benefits.
  - Stoke dangerous rivalries among countries and deepen fragmentation.
- Positive policy signals and processes already underway:
  - Legislation proposed by the EU that classifies AI by risk levels.
  - The Group of Seven Hiroshima AI Process.
  - The US executive order on AI.
  - The UK AI Safety Summit.
  - The UN's call for a high-level advisory body on AI as part of its Global Digital Compact.
- Suggested global mechanisms analogous to climate governance:
  - A shared framework similar to the Paris Agreement.
  - An expert panel modeled on the Intergovernmental Panel on Climate Change to track and share knowledge about AI.

### Role of international financial institutions (IFIs) and the IMF
- IFIs can help in three important areas:
  1. Gather knowledge at a global scale to better understand AI's broader effects on economies and societies, including through surveillance activities.
  2. Use convening power to provide a forum to share successful policy responses and build international consensus toward harmonized regulations.
  3. Bolster global cooperation via policy advice to ensure all countries reap AI benefits by:
     - Promoting the free flow of crucial resources such as processors and data.
     - Supporting development of necessary human and digital infrastructure.
     - Calibrating public instruments to support early-stage technologies without inducing fragmentation and restrictions across countries.
     - Encouraging public investment in AI and related resources while avoiding protectionism.

### Education, research, and labor market policies
- Labor market and social safety net recommendations:
  - Develop nimble social safety nets to help displaced workers.
  - Reinvigorate labor market policies to help workers remain in the labor market.
  - Assess taxation policies to ensure tax systems do not favor indiscriminate substitution of labor.
- Education and skills:
  - Demand for STEM specialists "will likely grow."
  - The value of a liberal arts education "may also increase" for training students to think across disciplines about big questions.
  - Education systems should be "at the frontier of AI development" by preparing the next generation and providing ongoing training for current employees.
- Public funding for AI research:
  - Historically, until 2014 most machine learning models came from academia, but industry has taken over.
  - In 2022, industry produced "32 significant machine learning models, compared with just three from academia."
  - Publicly funded AI research is needed to highlight the costs of AI to societies.

### Conclusions and forward-looking imperatives
- AI’s ability to mimic human thinking requires a unique set of global rules and policies to ensure societal benefit.
- The advent of AI demonstrates that "multilateral cooperation is more important than ever."
- Policymakers must "break out of our own echo chambers" and consider the broad interest of humanity.
- Scientific and technological progress "may be inevitable, but it need not be unintentional"; coordinated, responsible progress is required to achieve "a better life for more people."

*Harnessing AI for Global Good — Gita Gopinath, December 2023.*

---


_Source: https://www.imf.org/-/media/files/publications/fandd/article/2023/december/70-72-straight-talk-final.pdf_