Summary of Key Points
This interview focuses on the critical issues surrounding the development of AI globally: The current AI ecosystem is fragmented due to differences in values among countries, but this does not mean isolation. Instead, nations need to establish a relationship of strategic interdependence—each developing their own systems while sharing standards and resources. In the future, there may only be around 10 fundamental AI models left worldwide, so there's no point in wasting money on building new ones from scratch. The next breakthrough in AI will occur at the application level, not in basic models. Countries should identify their own areas of expertise (for example, the U.S. focusing on basic models, China on applications, and Europe on regulation). Governance requires a joint effort from governments, businesses, and technology companies; the core issue of security is defining the boundaries between humans and AI, not the technology itself.
Detailed Analysis
1. AI Fragmentation Does Not Mean Isolation
Why is there fragmentation? It's because different countries have different values—some place a high emphasis on privacy (e.g., the EU), while others prioritize efficiency (e.g., the U.S.). As a result, the design and rules of AI systems vary. However, fragmentation does not equate to isolation. The AI industry chain is too complex and costly (with $600 billion already invested globally, and an additional $400 billion expected annually); no single country can handle all aspects on its own.
What should be done? Basso suggests a strategy of strategic interdependence: each country can develop its own systems, but they must use common standards to ensure the smooth flow of data, talent, and funding. For instance, the World Economic Forum is promoting the concept of “digital embassies”—establishing physical entities in other countries while adhering to local laws, which allows for both sovereignty and cooperation. India recently revised its regulations to facilitate such exchanges.
2. The Competition for Fundamental Models: Focus on Applications
Fundamental models are the foundation of AI (such as GPT-4 and Wenxin Yiyan), but Basso believes that only about 10 will remain dominant in the future. Countries like China and the U.S. have already laid a solid foundation; investing heavily in new models might result in them becoming obsolete within a few years.
For example, rather than trying to develop its own fundamental model, Europe could use existing ones to build applications in healthcare or industry, which would be more cost-effective.
3. The Next Wave of AI Innovation Lies in Applications
Current models are already powerful (e.g., ChatGPT can write articles and perform analyses), but the real challenge is how to apply them in practical scenarios. Potential areas include:
- AI Agents: automating tasks like email management and scheduling;
- Embodied Intelligence: robots that assist with tasks or perform surgeries;
- Government Governance: using AI to monitor traffic and improve public services.
The country that successfully integrates AI into these areas first will gain a competitive advantage.
4. Identify Your Strengths and Focus on Them
Countries don't need to try to do everything; instead, they should focus on their areas of expertise:
- The U.S.: invest in basic models and data centers;
- China: excel in application development (e.g., AI-driven recommendations on platforms like TikTok or delivery services);
- Europe: focus on regulation and governance (e.g., developing AI-related laws);
- Morocco: cultivate top-tier technical talent.
Identifying and leveraging your strengths is far more effective than trying to be proficient in everything but failing to excel in any one area.
5. Governance and Security: Shared Responsibility with Clear Boundaries
Who is responsible for AI-related issues? It's not the sole responsibility of any one party:
- Application Companies: those using AI in medical diagnosis must implement security measures;
- Technology Companies: like OpenAI, which develop fundamental models, must ensure the security of their technology;
- Governments: they need to supervise and also use AI themselves (e.g., in smart cities).
Security issues are about how AI is used—not the technology itself. For example, when AI controls power grids, it's essential to define when humans should intervene (AI should not decide to cut off power supply on its own); when using AI for medical diagnoses, doctors must still provide final approval. Unclear boundaries can lead to risks.
The core message of this interview is clear: In the AI era, success does not depend on who develops the best fundamental models; it comes from finding one's place and collaborating with others to make AI a reality. For individuals and businesses, focusing on practical applications is more beneficial than chasing basic models. For countries, identifying and leveraging their strengths is crucial for success.