Research on the labor impact of GPTs is mainly focused on advanced economies in the West. Implicitly, it is assumed that what thrives in the former should also blossom in all other nations, provided they have reached a certain level of development and have been able to actively integrate into the global economy. The rest will simply miss the innovation train. However, such an assumption can be problematic, as it overlooks the current state of the international division of labor and disregards the unique characteristics of local contexts and cultures, as well as the long history of colonialism and postcolonialism.
Three critical issues must be highlighted at this point. To begin with, developing economies have distinct economic and social structures. Labor markets, socioeconomic organization, educational systems, and state capacity, among others, have distinctive features that, although varying from country to country, differ significantly from those of advanced capitalist countries.
Second, GPTs are not produced locally and thus must be imported and harnessed within national confines. That can entail significant capital investments, payment of IP royalties and recurrent usage fees, and deploying sophisticated human and institutional capacities. These capabilities are often not readily available. GPT diffusion and adoption could therefore be severely constrained, potentially having significant socioeconomic repercussions in the medium term.
Third, developing economies also face profound structural challenges. Gains in productivity in advanced nations due to AI and GPTs could reduce the need to import goods and services provided by developing nations. That, in turn, could impact the most dynamic sectors of those economies, slowing growth, curtailing job creation, and even freezing living standards in the best-case scenarios. Comparative and absolute economic advantages could vanish into thin air.
A critical research question here is how GPTs diffuse in developing economies, if at all. Undoubtedly, past GPTs such as the steam engine and electricity can provide significant evidence in this regard. For example, let us not forget that, two hundred years after its discovery, over 1 billion (1 x 10^9) people worldwide still have no access to electricity. Many with access have to contend with unstable voltage provisioning daily. In any case, time is crucial when discussing GPTs. Short, medium and long-term impacts must then be differentiated.
Historical evidence also shows GPT diffusion in core capitalist countries has led to steady automation and labor displacement, thereby triggering short-term job destruction. However, this has typically been complemented by medium— and long-term job augmentation in the most dynamic economic sectors and job creation in others. The rapid growth of high-skill jobs and the swift development of the service and financial sectors over the past 50 years, which have required a diverse range of skills, provides relevant evidence.
Economists and other researchers have developed at least three frameworks to explain the impact of pervasive technical change on jobs and skills.
The first, skilled-biased technical change (SBTC), argues that technological innovation disproportionately favors high-skilled individuals while simultaneously displacing low-skilled workers. In the short term, a premium is paid to those with advanced skills. The rest face a shrinking labor market and declining wages, or are forced to join the ranks of the unemployed. In the medium and long terms, wage inequality emerges as a distinct characteristic of widespread technical change. Computers and the Internet are good examples here. Jobs such as software programming and networking engineering became critical, while those requiring routine manual or intellectual inputs were mainly automated. At the policy level, strengthening technical and STEM education was identified as a top priority, accompanied by calls for making coding a standard at all education levels. However, existing empirical evidence does not fully support SBTC claims. Indeed, the framework cannot account for the increase in low-skilled jobs in the last 50 years or so, particularly in the rapidly expanding service sector. Moreover, SBTC cannot fully explain the labor markets in developing countries where underemployment, informal employment and unemployment are pervasive.
An alternative perspective based on the so-called tasks framework was developed to explain the growth of low-skilled jobs. Accordingly, sweeping technical innovation led to the elimination of medium-skilled routine tasks and jobs. Instead, it unleashed a new employment wave by promoting the employment of high and low-skill workers capable of performing non-routine (or non-repetitive) cognitive and manual tasks in the short term. In the long run, job polarization, accompanied by a U-shaped pattern of wage growth emerged. Workers with average skills who performed routine tasks had to either move up or down the job ladder or lose their jobs altogether. But like SBTC, this approach has little to say about the labor markets in developing economies. Moreover, job polarization cannot explain the potential automation of high-skill non-routine tasks and jobs, a purported key target of GPT AI. In any event, these two frameworks are not necessarily mutually exclusive. Indeed, both can be at work in many sectors of the economy today.
A third perspective emerges from classical political economy and heterodox economics. In brief, it connects the process of capital accumulation with the creation of a surplus population. It shares the SBTC assumption that technical change displaces low-skilled workers. However, it emphasizes that capital’s incessant drive for superprofits via technological innovation reduces the relative demand for labor, as automation and the ensuing productivity gains require more machinery and related physical inputs and proportionally less labor. A surplus working population thus emerges, surplus in relation to global capital itself, rather than to the overall population — or to excessive population growth, as propounded by Malthus. High-skilled workers are often rewarded but frequently experience layoffs when business cycles decline, only to be called back once they recover. The impact of capitalist innovation in agriculture creates a set of workers who migrate to cities and are readily available, as urbanization grows rapidly. A subset of this now large workers’ population, those with only basic skills, can hardly ever find employment and instead join the informal sector to make a living. The last cohort includes those who can work and manage to do so, avoiding capital’s tentacles, with others who live in extreme poverty and groups that engage in illegal activities altogether. While this approach does not address specific tasks, it appears to offer a more comprehensive framework for understanding the impact of technical change in developing countries.
In principle, GenAI and AgenAI could have skill-equalizing effects. Low-skilled workers can become more productive, while high-skilled workers may lose some of their financial perks, as demand for even more sophisticated skills increases. At the same time, the impact of GenAI and AgenAI could vary by sector. Sectors with higher cognitive and knowledge management requirements may be more affected than those involving complex, manual, non-repetitive labor skills. However, numeric AI, robotics, and ICTs will also impact the latter.
Unlike advanced Western nations, developing economies face exceptional challenges when confronting technical change driven by GPTs such as AI. An integrated approach that leverages the distinct positioning of each of them in the world economy is thus the way forward. Focusing exclusively on the impact of AI on jobs and skills alone will not suffice. Factoring in the country’s specific socioeconomic and cultural structure, which might be severely impacted by AI and other digital technologies, should be the starting point.
Raul
Selected References
Acemoglu, D., & Autor, D. (2011). Skills, Tasks and Technologies: Implications for Employment and Earnings. In Handbook of Labor Economics (Vol. 4, pp. 1043–1171). Elsevier. https://doi.org/10.1016/S0169-7218(11)02410-5
Acemoglu, D., & Restrepo, P. (2019). Automation and New Tasks: How Technology Displaces and Reinstates Labor. Journal of Economic Perspectives, 33(2), 3–30. https://doi.org/10.1257/jep.33.2.3
Arn, J. (1996). Third World Urbanization and the Creation of a Relative Surplus Population: A History of Accra, Ghana to 1980. Review (Fernand Braudel Center), 19(4), 413–443. JSTOR.
Arnold, D., & Pickles, J. (2011). Global Work, Surplus Labor, and the Precarious Economies of the Border. Antipode, 43(5), 1598–1624. https://doi.org/10.1111/j.1467-8330.2011.00899.x
Autor, D. H., & Dorn, D. (2013). The Growth of Low-Skill Service Jobs and the Polarization of the US Labor Market. American Economic Review, 103(5), 1553–1597. https://doi.org/10.1257/aer.103.5.1553
Autor, D. H., Levy, F., & Murnane, R. J. (2003). The Skill Content of Recent Technological Change: An Empirical Exploration. The Quarterly Journal of Economics, 118(4), 1279–1333. https://doi.org/10.1162/003355303322552801
Berman, E., Bound, J., & Machin, S. (1998). Implications of Skill-Biased Technological Change: International Evidence*. The Quarterly Journal of Economics, 113(4), 1245–1279. https://doi.org/10.1162/003355398555892
Bernards, N. (2018). The global governance of precarity: Primitive accumulation and the politics of irregular work. Routledge.
Brynjolfsson, E., Rock, D., & Syverson, C. (2018). Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics. In NBER Chapters (pp. 23–57). National Bureau of Economic Research, Inc. https://ideas.repec.org/h/nbr/nberch/14007.html
Cantens, T. (2023). How Will the State Think With the Assistance of ChatGPT? The Case of Customs as an Example of Generative Artificial Intelligence in Public Administrations (SSRN Scholarly Paper No. 4521315). https://doi.org/10.2139/ssrn.4521315
Cazzaniga, M., Jaumotte, F., Li, L., Melina, G., Panton, A. J., Pizzinelli, C., Rockall, E. J., & Tavares, M. M. (2024). Gen-AI: Artificial Intelligence and the Future of Work. Staff Discussion Notes, 2024(001). https://doi.org/10.5089/9798400262548.006.A001
Chun, J. J. (2016). The Affective Politics of the Precariat: Reconsidering Alternative Histories of Grassroots Worker Organising. Global Labour Journal, 7(2). https://doi.org/10.15173/glj.v7i2.2483
Ernst, E., Merola, R., & Samaan, D. (2018). The economics of artificial intelligence: Implications for the future of work (p. 41). ILO.
Freeman, C., & Louçã, F. (2002). As time goes by: From the industrial revolutions to the information revolution. Oxford Univ. Press. https://doi.org/10.1093/0199251053.001.0001
Freeman, C. (1995). The ‘National System of Innovation’ in historical perspective. Cambridge Journal of Economics, 19(1), 5–24. https://doi.org/10/gdk2vr
Goldin, C., & Katz, L. F. (2009). The Race Between Education and Technology (First Harvard University Press paperback edition). The Belknap Press of Harvard University Press.
Griliches, Z. (1969). Capital-Skill Complementarity. The Review of Economics and Statistics, 51(4), 465. https://doi.org/10.2307/1926439
Katz, L. F., & Murphy, K. M. (1992). Changes in Relative Wages, 1963-1987: Supply and Demand Factors. The Quarterly Journal of Economics, 107(1), 35–78. https://doi.org/10.2307/2118323
Krusell, P., Ohanian, L. E., R�os-Rull, J.-V., & Violante, G. L. (2000). Capital-Skill Complementarity and Inequality: A Macroeconomic Analysis. Econometrica, 68(5), 1029–1053. JSTOR. https://doi.org/10.1111/1468-0262.00150
Neilson, D., & Stubbs, T. (2011). Relative surplus population and uneven development in the neoliberal era: Theory and empirical application. Capital & Class, 35(3), 435–453. https://doi.org/10.1177/0309816811418952
Perez, C. (2003). Technological revolutions and financial capital: The dynamics of bubbles and golden ages (Repr). Elgar.
Pintera, J. (2025). Skill-bias and wage inequality in the EU New Member States: Empirical investigation. Structural Change and Economic Dynamics. https://doi.org/10.1016/j.strueco.2025.05.024
Shao, Y., Zope, H., Jiang, Y., Pei, J., Nguyen, D., Brynjolfsson, E., & Yang, D. (2025). Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce (No. arXiv:2506.06576). arXiv. https://doi.org/10.48550/arXiv.2506.06576
Tu, X., Zou, J., Su, W. J., & Zhang, L. (2023). What Should Data Science Education Do with Large Language Models? (No. arXiv:2307.02792; Version 1). arXiv. https://doi.org/10.48550/arXiv.2307.02792
