The Power and Perils of the “Artificial Hand”: Considering AI Through the Ideas of Adam Smith
IMF News, June 5, 2023
Source details
- Canonical URL
- The Power and Perils of the “Artificial Hand”: Considering AI Through the Ideas of Adam Smith
Other formats
Bibliographic details
- Authors: Gita Gopinath
- Published: June 5, 2023
Introduction
- Speech by Gita Gopinath, First Deputy Managing Director, IMF.
- Occasion: Speech to commemorate 300th anniversary of Adam Smith’s birth, University of Glasgow, June 5, 2023.
- Framing: Uses Adam Smith’s ideas (The Wealth of Nations; The Theory of Moral Sentiments) as a lens to assess generative artificial intelligence (AI), its economic potential, labor-market implications, concentration risks, regulatory needs, and societal effects.
Productivity and growth: potential gains and evidence
Findings:
- AI could raise productivity by automating certain cognitive tasks and giving rise to new higher-productivity tasks for humans.
- Example empirical result: a study of customer-service agents using a conversational assistant based on generative AI found productivity rose by 14%.
- Goldman Sachs forecast: AI could increase global output by 7%, or roughly $7 trillion, over a decade.
- Historical parallel: The Industrial Revolution—Smith’s era—similarly transformed work and productivity.
Labor-market impacts and distributional concerns
Findings:
- Automation previously has led to loss of “middle-skill” jobs and polarization into high-paying and low-paying clusters.
- Recent empirical studies suggest AI could affect occupations and industries differently than prior automation waves and may put downward pressure on wages of high-paying jobs.
- Some studies indicate AI adoption could flatten firm hierarchies: increasing the number of workers in junior positions and decreasing the number in middle management and senior roles.
- Scale of vulnerability: some researchers estimate that two-thirds of U.S. occupations could be vulnerable to some form of automation.
- Uncertainty: net impact on employment is not guaranteed to be positive—AI might replace human jobs without creating new, more productive work for humans to transition into (citing Daron Acemoglu’s concern).
Market concentration, data, and compute power
Findings:
- The market for components to develop AI tools is highly concentrated (example cited: a single company dominant in silicon chips best suited for AI applications).
- Many AI models require massive computing power and huge amounts of data, concentrating capability among a handful of large corporations despite open-source efforts.
New approach to regulation and public policy (policy recommendations)
Policy recommendations:
- Urgently develop sound, smart regulations to ensure AI is harnessed for societal benefit.
- Recognize that AI may require an entirely new game and an entirely new approach to public policy.
- Support regulatory models that classify AI by risk levels (example: EU’s Artificial Intelligence Act):
- Highest-risk systems banned (includes government systems that rank people based on social compliance, “social scoring”).
- Next-highest risk level tightly regulated with requirements for transparency and human oversight.
- Address broader economic and social effects of AI:
- Develop nimble social safety nets for those whose jobs are displaced.
- Reinvigorate labor market policies to help workers remain in the labor market.
- Carefully assess taxation policies to ensure tax systems don’t favor indiscriminate substitution of labor.
- Education and training:
- Adjust education to prepare the next generation to operate new technologies and provide ongoing training for current employees.
- Anticipate increased demand for STEM specialists while recognizing increased value of liberal arts education for interdisciplinary thinking.
- International coordination:
- Coordinate regulation across borders; the G7 has formed a working group to study AI.
- Aim for a truly global set of rules given the cross-border nature of AI and the speed of technological change.
Moral sentiments, information integrity, and societal risks
Findings and concerns:
- Adam Smith’s concept of “sympathy” underpins moral behavior and a rules-based society; generative AI’s mastery of language could erode these social bonds.
- Generative AI can comb vast knowledge and produce convincing messages; it can “speak like us,” raising risks of indistinguishability between human and machine interlocutors.
- Risks noted:
- AI models can replicate embedded biases from training data.
- AI “hallucination”: models confidently defending false information.
- “Fake intimacy” (per Yuval Harari): AI forming close relationships with people to influence opinions and worldviews, potentially destabilizing societies and undermining accepted social narratives.
- Existential concerns: more than 350 AI industry leaders signed a statement calling for global priority to mitigate the risk of “extinction” from AI, placing it on par with pandemics and nuclear wars.
- Information integrity:
- AI could damage the integrity of information flows—market price signals and emotional cues—weakening the foundations Smith described for functioning markets and civilized behavior.
- Suggested protective measures:
- Support rules that protect consumer privacy and limit misinformation in the age of AI.
- Maintain human oversight in critical areas (medicine, critical infrastructure) to mitigate severe risks.
Conclusion
- AI could be as disruptive as the Industrial Revolution; balancing innovation support with regulatory oversight is essential.
- Because AI can mimic human thinking and speech, it demands a unique set of rules and global coordination.
- Navigating AI’s power and perils will require interdisciplinary approaches, broad empathy, and ingenuity—drawing on law, history, rhetoric, languages, mathematics, and economics—as Smith exemplified.
- The debate is ongoing; no definitive answers were claimed, but a set of priorities and concerns were presented to guide policy and societal responses.
Source: Speech — Gita Gopinath, "The Power and Perils of the “Artificial Hand”: Considering AI Through the Ideas of Adam Smith", University of Glasgow, June 5, 2023.