Artificial intelligence is emerging as a powerful force reshaping the global order. From the reorganization of industrial supply chains to the production of knowledge, from transformations in the nature of work to international competition, the impact of AI has already gone far beyond a mere technological revolution. It is becoming a question of the very structure of the societies of the future.
When large corporations control the most advanced AI technologies, could this give rise to a new form of colonialism toward smaller countries that lack such technological capabilities? Are the data generated by people in Africa truly equal, in the eyes of AI, to the data generated in Europe and the United States? Whose intelligence? Who, ultimately, does artificial intelligence serve? Is it creating a more inclusive and equitable future, or is it reproducing, and perhaps deepening, existing inequalities in new technological forms?
In this conversation, Professor Yin Zhiguang of Fudan University speaks with Hansong Li, Assistant Professor at American University and Research Fellow at the Harvard Kennedy School’s Center for International Affairs, about AI from the perspective of the Global South， examining the global division of labor, data power, and knowledge justice behind the technology.
Why are you interested in AI, and what are your main concerns about its impact on society and the global order?
AI is an unavoidable question of our time because it represents one of today’s most advanced productive forces. But we should neither assume that technology follows a single, linear path nor believe that AI can automatically solve all social problems. Its significance lies not only in transforming production, but also in reshaping social relations and collective consciousness.
I think we can understand AI on at least three levels. The first is the restructuring of global supply chains and the international division of labor: why do some countries and companies control large language models, computing power, and core algorithms, while others are left providing data, labor, minerals, and energy? How did this new division of labor emerge?
How does the future envisioned by large data companies like Palantir differ from our existing forms of social organization, and does it truly promote equality or instead reproduce inequality?
If AI is truly empowering us, is that empowerment universal and equal, or does it accelerate and amplify existing inequalities? This raises a deeper question: What will the AI utopia of the future actually look like?
Many visions promoted by tech leaders appear human-centered, but are in fact centered on a technological elite, surrounded by robots while ordinary people remain outside. In the Global South, this raises an especially urgent problem: as automation advances, many workers may never experience the kind of industrialization that transformed East Asia, nor gain access to the high-end technology sector. Such a future could therefore reproduce profound structural inequalities.
A second vision is human–machine integration, in which AI becomes an extension of human capabilities, a kind of “second brain” that gives us instant access to knowledge and information. But this does not automatically solve social problems. It may increase individual productivity while simultaneously creating new forms of labor, social relations, and alienation. Ultimately, how humans integrate with machines is not simply a technological question; it is a question of political economy and the kind of society we choose to build.
To what extent does AI-enhanced human–machine integration actually increase productivity across different forms of labor, and how should we understand its impact when AI becomes an extension of human capabilities rather than merely a tool?
AI may increase efficiency without necessarily increasing social productivity as a whole, and its impact remains highly uneven across sectors. In developed economies, service-sector jobs such as legal assistants, accountants, financial analysts, and programmers are already being significantly reshaped, while the rise of AI agents and human–machine integration could make the future of human labor increasingly uncertain.
By contrast, agriculture, small-scale production, and the informal economy have a much more complicated relationship with AI. In countries like India, where a vast informal economy exists alongside a rapidly growing AI sector, it is still unclear how deeply AI is actually transforming the lives of ordinary workers. At the same time, the Global South also offers alternative possibilities, such as using AI for weather monitoring, telemedicine, and agricultural services. The question, then, is not simply whether AI increases productivity, but whose productivity it increases, and under what conditions.
Does AI create new forms of inequality, or does it primarily deepen existing disparities between agricultural and industrial countries and between the primary and tertiary sectors?
Both dynamics are happening at the same time. On the one hand, AI is creating new inequalities within its own global supply chain. Countries at the bottom often perform data labeling and content moderation, labor-intensive and psychologically damaging work involving violent, sexual, or otherwise disturbing content. They effectively become the “filters” of the AI industry, absorbing the risks and trauma while others receive the clean data. This echoes older patterns of global production, in which developing countries have long borne the environmental and human costs of global consumption.
On the other hand, AI empowers countries, regions, and social groups very unevenly, creating new forms of dependence and exclusion. Some become the “front of the train,” while others are left behind, or never get on the train at all. When the knowledge economy fails to be inclusive, those excluded may understandably ask, “What does AI have to do with me?” That sense of exclusion can fuel resentment toward technological elites and, ultimately, populist politics.
Given the deep inequalities within and beyond the AI industry, how can we rethink global AI governance to account for these differences and build a more equitable framework?
This goes beyond inequality in the AI supply chain and raises a broader question of global knowledge injustice. Inclusive AI governance must begin with inclusive data. Although AI models are trained on enormous datasets, they still capture only a small fraction of human experience. What gets included as “valid data” is itself a political and epistemic choice, largely shaped by major technology companies and engineers in the Global North.
That is why we need to bring more countries, languages, cultures, and forms of local knowledge into AI development. We can already see this in Africa, where startups in countries such as Uganda and Ethiopia are developing models based on local languages, data, and social realities rather than simply reproducing existing English-language models. This shows that AI does not have to follow a single Silicon Valley model.
Ultimately, this is a question of epistemic justice: Whose data enters the model? Whose knowledge is represented? And who remains excluded from knowledge production? If AI governance is to be genuinely inclusive, we must diversify not only the languages of AI, but also the data, knowledge, and perspectives that shape it.
The third level concerns our imaginaries of the future. Some technology elites in Silicon Valley envision futures in which a small minority controls AI, embodied robots, and other advanced technologies—or even escapes to Mars while the rest of humanity is left behind. Such visions point toward an increasingly technocratic and hierarchical future, with echoes of what has been called the Dark Enlightenment.
What are the material conditions required for AI to become truly accessible and inclusive worldwide, from affordable and open models to low-cost, reliable energy?
Beyond open source, AI needs to be meaningfully integrated into social life and economic production. We could even slow the race to push the limits of generative AI and large language models, and focus more on what they actually contribute to ordinary people and society.
A second priority is sustainable energy. AI depends heavily on electricity, so a diverse and sustainable energy system is itself a foundation for AI development. This is also why energy security and AI governance cannot be treated as separate issues.
Ultimately, however, the most important question is where society wants AI to go. Countries and regions such as Singapore, ASEAN, and the African Union are already exploring new approaches to AI governance, while the United States has traditionally relied more on market-driven innovation, intervening only after problems become serious. This reflects a broader tendency to pursue technological progress without first building the social foundations necessary to sustain it.
We need to bring human needs and lived realities back into discussions of technological development and global governance. Technology cannot be an end in itself; it must be grounded in the material and social conditions of human life. Otherwise, we risk building a technological “second floor” while forgetting that the foundations beneath it are already cracking.
How can we put human needs and lived realities back at the center of technological development and global AI governance, rather than assuming that technological progress alone will solve broader social problems?
The key is collaborative governance. For too long, global resources and decision-making power have been concentrated in the hands of a few countries and corporations, leaving many developing countries with little voice in the rules that govern them. As the old slogan goes, “no taxation without representation.” The same principle should apply to AI governance: countries should not simply be expected to accept rules they had no role in creating.
The most fundamental principle of global AI governance, therefore, should be co-creation and shared governance. We need to develop a shared wisdom around AI, build consensus through participation, and gradually establish norms that everyone has a stake in. These norms will inevitably evolve, but through continued cooperation they can develop into genuinely universal mechanisms of governance.
This also creates an opportunity for the Global South to move beyond being merely a recipient of supposedly “universal” rules. By participating in the construction of those rules, the Global South can become a co-creator of what universality itself means.
Source: guancha, live science, techradar
