Summary of the Key Points
This speech uses Serbia as a case study to explore the fundamental debate surrounding the concept of "AI inclusivity": should 8 billion people become users of AI products owned by a handful of multinational corporations, continuing the "digital colonization" logic of neoliberal globalization, or should more countries become producers, innovators, and distributors of AI technology? The speech analyzes the challenges faced by developing countries in the AI era, such as the "brain drain" and "value chain lock-in," and divides the countries of the Global South into three tiers of AI development. It also discusses how China-Serbia cooperation can avoid becoming a form of "digital colonization." China's approach to integrating AI technologies requires deep localization to foster local talent and industrial growth.
Two Paths to AI Inclusivity: "Users" or "Producers?"
When we talk about "AI inclusivity," there are actually two distinct paths:
- Path A: Enabling more people to become users—perhaps by making ChatGPT or similar tools accessible to 8 billion people worldwide. However, this essentially represents a new form of neoliberalism, where a few developed countries (like the United States) control AI research and development, standards, and platforms, while developing countries provide only markets, data, and talent, with all profits and technological control remaining in the hands of these foreign entities. This is similar to how colonies used to supply raw materials to their colonial masters; today, it's about selling data and intellectual labor, resulting in a situation where developing countries are trapped at the bottom of the value chain—what could be called "digital colonization."
- Path B: Encouraging more countries to become producers by enabling them to develop their own AI technologies (creating local models and training their own engineers), thereby gaining control over intellectual property and rule-making. Only in this way can AI truly help developing countries boost their productivity and avoid becoming mere profit sources for others.
The speech emphasizes that if only Path A is pursued, AI inclusivity will be a mere facade; only by following Path B can developing countries break free from this controlling fate.
Brain Drain: From an "Open Game" to a "Hidden Trap"
In the pre-AI era, brain drain was a clear issue—talented engineers would directly immigrate to Europe and America or return home after studying abroad. But in the AI era, the situation has become more subtle:
- People stay at home, but the benefits go abroad: For example, Microsoft's research center in Serbia employs 800 engineers who live locally, yet the code they write, the algorithms they develop, and the patents they apply for all belong to Microsoft. These talents contribute value to a foreign company while leaving developing countries with only jobs and limited tax revenue.
- Even worse than physical migration: While manufacturing plants require local supply chains, AI research does not; code can be uploaded to the cloud with a click, and the resulting intellectual assets belong to multinational corporations. Developing countries end up with only the jobs and some tax revenue, while the core technologies and profits go abroad.
Three Tiers of AI Development for the Global South
The speech categorizes developing countries into three tiers:
1. Tier 1: Sovereign AI builders—countries with the capital, policies, and computational resources to develop their own AI models (e.g., the UAE, India, Brazil). They compete for the status of regional AI hubs, attracting multinational corporations and developing strong local models.
2. Tier 2: Countries with talent but lacking key elements—such as Serbia. These nations have skilled engineers and use cases (e.g., in healthcare and agriculture) but lack large domestic companies, affordable computing power, and a technology ecosystem for smaller languages. Their challenge is to retain the value created by their talent.
3. Tier 3: Pure users—countries like the Democratic Republic of Congo and Mozambique, which do not even have widespread internet access and can only use pre-made AI products (like ChatGPT). They must first address basic infrastructure issues before considering AI integration.
Serbia falls into Tier 2; it has a foundation but struggles with retaining the value created by its AI efforts.
Serbia's Advantages and Challenges
Serbia's strengths include:
- A comprehensive AI strategy: The country adopted an AI strategy in 2019, with a new plan for 2025-2030 covering legal frameworks, talent development, and research.
- A strong talent base: Microsoft has expanded its research center in Serbia from 8 to 800 employees, indicating a large pool of skilled professionals.
- Access to data centers and use cases (e.g., in healthcare and agriculture).
However, Serbia also faces challenges:
- Lack of domestic companies that can drive the AI industry (such as Huawei or Microsoft).
- Insufficient computing power: The cost of GPUs, essential for AI, is prohibitive.
- A weak technology ecosystem for smaller languages: Although Serbian is spoken by 20 million people in neighboring Slavic countries, there is a lack of suitable adaptations.
Serbia's strategic focus is noteworthy: **laws → talent → research → use cases → data → computing power.* Why not invest in GPUs first? Because without local companies and talent, computing power would only benefit foreign entities, increasing dependence. The correct order is to establish legal frameworks (in line with EU standards), retain talent, and then build computing capacity.
China-Serbia Cooperation: Why It's Less Likely to Be a Form of "Digital Colonization"
Chinese companies' approach to AI exports focuses on an integrated model that includes open-source models, hardware, energy, and specific use cases. For example, Chinese security systems must be adapted for Serbian languages and algorithms to recognize local facial features. This requires local engineering involvement, ensuring that technical knowledge remains in Serbia. China can also provide a holistic solution for infrastructure (e.g., nuclear power generation, data centers, AI applications) rather than just selling individual components.
This approach differs from that of American companies, which typically sell software or platforms with minimal local participation. Chinese cooperation with Serbia can help Serbia develop its own AI capabilities, preventing digital colonization.
The True Meaning of AI Inclusivity
The essence of AI inclusivity is not about ensuring everyone has access to AI tools but about enabling everyone to participate in AI innovation. Developing countries must move from being users to producers by retaining talent, building local ecosystems, and mastering core technologies. China's cooperation model offers a new option for them: they don't have to rely on Western multinational corporations but can take control of their own AI future through deep collaboration.
In the words of the speech, "Computing power is a magnifier, not a substitute. Without your own seeds (local companies and talent), the magnifier will only amplify dependency; with them, you can grow your own trees." This is the true direction of AI inclusivity.