Overview
Tom Friedman discusses the future of artificial intelligence (AI) and argues against the prevailing US-China "AI race" mindset, urging instead for unprecedented collaboration to prevent destabilization and misuse. The conversation highlights the urgency of embedding ethical controls in AI and the mutual benefit of cooperation between the US and China.
Spectacle of US-China-India-Russia Relations
- Recent public alignment among India, Russia, and China is viewed as spectacle, reflecting American diplomatic failure.
- Despite the display, deep military or strategic alliance between China and India remains unlikely due to historic tensions.
- American isolation increases as it loses diplomatic leverage when traditional allies, like India, move closer to rivals.
Urgency and Dangers of Autonomous AI
- AI and climate change are rapidly advancing, with AI soon approaching superintelligence.
- AI is distinct as a "vapor" technology, integrating into every device, making control complex and essential.
- Unlike previous technologies, AI is not just dual-use but "quadruple use," potentially making autonomous harmful decisions.
- The emergence of AI as a new, non-human agent poses unprecedented ethical and safety risks.
The Case for US-China AI Collaboration
- Controlling AI requires US-China cooperation to embed shared ethical and legal norms within every AI device.
- Failure to collaborate risks global autarky, diminished trade, and distrust of foreign technology (e.g., AI-infused medical devices).
- A joint "AI adjudicator" could ensure AI alignment with the laws and normative values of each country.
- Early experiments show potential in training AI with fables to foster moral reasoning.
Practical Proposals for Building Trust
- Establish a technical substrate ("AI adjudicator") in all AI devices to filter decisions according to legal and ethical standards.
- US and China set shared rules, then require global trading partners to adopt the same standards for economic engagement.
- Admits likely political barriers, especially amid heightened US-China antagonism, but stresses lack of alternatives.
Limitations of Third-Party Regulation
- EU-style regulation alone is insufficient without a US-China core agreement.
- The rest of the world would have to adopt US-China standards to participate in major economic and technology platforms.
Risks of AI Misuse and Rogue Behavior
- AI systems can already exhibit dangerous, unexpected choices in simulation scenarios.
- Designers' limited understanding of AI emergent behaviors increases risk.
- Deepfakes and AI-enabled fraud could destabilize societies, start conflicts, and empower bad actors.
Differentiation from Past Technologies
- Unlike nuclear weapons, which remained government-controlled, AI proliferation is widespread and accessible.
- Lack of controls could repeat or exceed the societal disruption seen from unregulated social media.
Recommendations / Advice
- Avoid the mindset of an "AI race"; instead, foster US-China trust and technological collaboration.
- Develop and deploy technical and diplomatic safeguards to align AI with human well-being.
- Proactively regulate AI to prevent severe social, political, and security harms before widespread adoption.