by István Kopácsi
The report "Mind the AI Divide: Shaping a Global Perspective on the Future of Work"[1] (Report) co-authored by the United Nations and the International Labour Organization (ILO) delves into the profound impact that Artificial Intelligence (AI) is having on economies, societies, and labour markets across the globe. AI's rapid advancement is not affecting all regions equally, leading to what is termed an "AI divide" that exacerbates existing disparities between developed and developing nations. While high-income countries are positioned to benefit significantly from AI innovations, low- and middle-income countries, particularly in Africa, face the risk of falling further behind due to limited digital infrastructure and a shortage of relevant skills. If proactive international cooperation and targeted policies are not implemented, AI’s potential to drive sustainable development and alleviate poverty will remain unfulfilled. The labour market, where AI's impact on productivity and job quality is most apparent, exemplifies this divide.[2]
Uneven Ground: AI's Role in Reshaping Labour Markets
AI's capacity to reshape labour markets is undeniable, but its effects vary greatly across regions and occupations. Research from the ILO indicates that while certain jobs, especially in clerical roles, are highly susceptible to automation, the fear of massive job losses is exaggerated. Instead, AI is likely to partially automate tasks, leading to improved efficiency and job augmentation rather than outright job elimination.
Clerical support workers, for instance, are among those most exposed to automation, with a significant portion of their tasks at risk. In other professions, however, only a small percentage of tasks face high exposure to automation. Geographically, the share of employment at risk of automation is highest in Europe and North America due to their diversified economies. In contrast, regions such as Latin America, Asia, and Africa exhibit lower exposure because of a higher concentration of jobs in sectors like agriculture, which are less prone to AI-driven automation. Furthermore, gender disparities are evident, with women, particularly in clerical roles, facing higher exposure to automation. This could disproportionately affect them, especially in sectors such as business process outsourcing, which are crucial to the economies of developing countries.
Despite these automation risks, there is also considerable potential for AI to augment jobs, especially in developing regions, provided there is an improvement in digital infrastructure and skills. Countries like Kenya and Rwanda have demonstrated how technological advancements, such as mobile internet, can spur job creation and economic growth, showing that AI can bring significant benefits when the right conditions are in place.[3]
While AI can enhance productivity, its integration into the workplace raises important concerns about job quality and working conditions. The rise of algorithmic management, where algorithms optimize and monitor work, presents both opportunities and challenges. On one hand, it can increase efficiency; on the other, it can reduce workers' autonomy and intensify work pressure. Algorithmic management is already prevalent in digital labour platforms and offline industries like warehousing, where technology dictates work tasks and pace, often limiting workers' influence over their work environment. The overall impact of AI on job quality largely depends on the extent to which workers can participate in decisions regarding the integration of technology. Countries with strong traditions of worker participation, such as the Nordic countries and Germany, tend to have more positive outcomes in terms of AI adoption in the workplace. Ensuring that workers have a voice in how technology is implemented and used is crucial for AI to positively influence job quality.[4]
The AI Value Chain and the Demand for Skills
The AI value chain, much like other global value chains, is characterized by distinct stages that each require specific skills and infrastructure. The value derived from these stages varies widely, with high-income countries dominating the high-value stages, such as design and deployment, while lower-value activities are more common in developing countries. The AI value chain begins with data collection, which serves as the raw material for AI systems. This data comes from various sources, including publicly available information and proprietary datasets, and often requires data labelers and annotators, typically low-skilled workers in developing countries. As we move further along the value chain, model design, training, tuning, deployment, and maintenance differ significantly from data annotation, requiring more physical infrastructure and advanced expertise from computer scientists or STEM graduates, along with substantial R&D investments. Finally, AI systems must be continuously updated and maintained, a stage also dominated by more developed regions.
The distribution of skills and value in the AI value chain highlights disparities in how different countries can benefit from AI. Developing countries often participate in the lower-value stages, while developed nations control the high-value aspects, exacerbating the global AI divide. The demand for AI-related skills is growing rapidly, but there remains a significant skills gap, particularly in developing countries. Bridging this gap requires not only technical education but also a broader focus on building the infrastructure necessary to support AI development.
The Report highlights the importance of skills development in addition to capital investments in AI for developing countries to fully benefit from the technology. Skills are crucial not only for creating national AI systems but also for understanding and using these systems effectively, including instituting safeguards. In the workplace, companies adopting AI must consider its implications for cybersecurity, data privacy, and potential discrimination, especially in recruitment. Skills training can enhance the integration of AI by informing managers and workers about data use, worker autonomy, and feedback mechanisms.
A worker-centric approach is recommended to evaluate AI's impact, focusing on both the technical aspects and broader social implications. This approach involves analyzing how AI affects specific tasks and skills, enabling the identification of sectors most vulnerable to automation. A well-structured lifelong learning framework is essential for developing targeted skills programs to help workers adapt to AI-driven changes. [5]
Moving Forward: Strengthening International Cooperation, Building National Capacity, and Addressing AI in the World of Work
To address the challenges posed by AI, the Report advocates for strengthened international cooperation, national capacity building, and a focus on AI's role in the world of work. International cooperation is essential, with developed countries bearing a responsibility to support AI capacity building in developing nations. This support could include financial aid, knowledge sharing, technology transfer, and collaborative research. At the national level, countries need to develop strategies tailored to their specific contexts, focusing on building the infrastructure and skills necessary to harness AI for economic and social development. Ensuring that AI contributes positively to the world of work requires a focus on job quality, worker participation, and social dialogue. This approach can help mitigate the potential negative impacts of AI, such as job displacement and worsening working conditions.[6]
Conclusion
The future of work in an AI-driven world will depend on how well countries can navigate the challenges of AI adoption. By focusing on cooperation, capacity building, and a fair distribution of AI's benefits, the Report suggests it is possible to create a more equitable global AI ecosystem that supports sustainable development and shared prosperity.
[1] https://www.ilo.org/publications/major-publications/mind-ai-divide-shaping-global-perspective-future-work
[2] Report, 3.
[3] Ibid., 6-9.
[4] Ibid., 10.
[5] Ibid., 11-14.
[6] Ibid., 17-19.
