Summary of Key Points
The global AI competition is shifting from a resource-intensive approach focused on having the largest models, most computing power, and the highest number of parameters to a more resource-friendly model that enables more countries to access AI capabilities at low costs. For the countries in the Global South, which struggle to replicate the Silicon Valley model (developing nations), the four panelists proposed a development strategy that avoids an arms race and emphasizes local adaptation. They argue that AI should be regarded as a national public infrastructure rather than a commercial product. It is crucial to invest in local knowledge accumulation rather than blindly pursuing large models, to prevent the loss of talented individuals, and to build a domestic innovation ecosystem. The ultimate goal of AI should be to create new production capabilities rather than merely integrating it into existing industries.
1. AI Development: Stop Focusing on Parameters and Ask Whether It Is Useful
In the past two years, the AI community has been engaged in an almost arms race, with companies like OpenAI and Google competing to develop models with the largest number of parameters and the highest rankings. Professor Shen Yi from Fudan University raised a poignant question: “Can ordinary users tell the difference between a score of 88 and 86 on a ranking?” He pointed out that AI is not about competing for high scores but about solving real problems—helping farmers improve their farming methods, increasing efficiency in factories, and making healthcare more affordable. Just as water, electricity, and telecommunications are national infrastructures, AI should also be a public asset linked to education, industry, and governance, rather than being the private property of a few tech companies.
More importantly, countries in the Global South should not be forced to choose between giving up their sovereignty to use someone else’s AI or having no means to develop their own. This “kneeling to make money” option is unethical. Every country should have the right to decide how to use data and deploy models; this is what truly represents “sovereign equality” in the AI era.
2. Knowledge Sovereignty: It Is More Important to Retain Local Experience Than to Have Our Own Models
Ivana, a scholar from Venezuela, noted that the challenges for countries in the Global South remain the same: during the colonial era, they exported minerals; in the industrial era, they exported resources; now, they export data—still relying on others. For example, although India has many data centers, its cloud computing market is controlled by foreign companies. She believes that true “digital sovereignty” means having control over one’s own knowledge. This involves collecting local expertise (such as how farmers grow crops or how communities solve problems) and turning it into a knowledge base. In Brazil, for instance, a project organized farmers’ oral farming experiences into a database and then used open-source models to develop agricultural AI. The model is not the focus; what matters is who determines which knowledge is important.
If we do not collect this local knowledge now, future AI models will not reflect our own perspectives, and we will be dependent on others’ “universal knowledge” to solve our problems.
3. A New Form of Talent Loss: People Stay, but Knowledge Leaves
Filip from Serbia discussed the phenomenon of “hidden immigration,” where talent moves abroad despite staying in their home countries. Companies like Google and Microsoft set up research centers in developing nations, hiring top engineers, but the resulting innovations and intellectual property often belong to foreign companies, with little benefit for the host countries. Serbia’s AI strategy prioritizes legal governance and talent development over just building computing power. Without a local ecosystem, even more advanced technology (such as GPUs) will only increase dependence on foreign platforms. It is more urgent to create an environment that retains talent and knowledge.
4. AI No Longer Searches for Use Cases; It Creates New Production Capabilities
Zhao Zhongxia, from the industry, explained that in the past, AI was like using a hammer to find nails—trying to apply facial recognition to all industries or redefining cars with autonomous driving. However, large models are different now; they directly create new capabilities. For example, AI can execute tasks on its own (agents) and write code, and multiple AI systems can collaborate to form new work structures. He cited the example of how employees’ skills can be acquired by AI, turning them into company assets. In the future, workers will sell not just their time but their skills. This means that AI will not replace jobs but reshape the entire production system.
Conclusion: The Opportunity for the Global South Is to Follow Its Own Path
The panelists all agree that the ultimate goal of AI competition is not about who has the largest model but about transforming AI into a country’s own development engine. Chips and computing power are important, but they are just tools. What really makes a difference is whether a country can retain local knowledge, leverage talent for its own benefit, and use AI to solve its own problems. Instead of copying Silicon Valley, countries in the Global South should find their own path—perhaps by using open-source models combined with local knowledge bases to serve agriculture or developing small, practical AI tools that become accessible to everyone. This is the true meaning of making AI inclusive.