AI will not save the South

Artificial intelligence technology has been held out as a solution to many of the challenges facing the developing countries but, given the current functioning of the digital economy, may only end up reinforcing their marginalisation.

Cédric Leterme

---

‘IMAGINE you are the agriculture minister of a developing country tasked with quickly identifying the cause of leaf damage across a number of farms in order to detect the presence of pests that could threaten your country’s food security. … Artificial Intelligence (AI) is the engine currently driving innovative solutions towards tackling these types of problems, and the faster governments can support and adopt AI as part of a broader digital strategy, the better positioned they will be to quickly respond to their own development challenges.’

This is an excerpt from a World Bank report entitled ‘Harnessing artificial intelligence for development’.¹ In a typically ‘solutionist’ vein,² this institution – and many others – strives to show how AI could be put to the service of development, provided that governments of the South adopt the ‘right approach’ to ‘maximise opportunities and limit risks’. In doing so, not only do these organisations depoliticise fundamentally sociopolitical issues by reducing them to narrowly technical dimensions that can be automatically processed. But above all, they ignore the numerous structural threats that current developments in AI pose to the countries of the South.

A double economic and geographical concentration

These threats become clearer if we consider AI in relation not only to its effects but, first and foremost, to its production conditions. This was a position advocated by the philosopher Nick Srniček in a recent interview: ‘Beyond the media hype, what I propose is to make a classic Marxist gesture: rather than focusing on fears and consequences of the use of artificial intelligence, we must take an interest in its production conditions.’

In doing so, we can then relate to this observation made by Indian digital expert Anita Gurumurthy: ‘The AI-led economy as we know it, is not an accident. From the relatively innocent Internet of the 90s through Snowden, and the rise and rise of the FAANG [Facebook, Amazon, Apple, Netflix and Google], to Cambridge Analytica, we have seen the unfolding of a data culture that is deeply intertwined with capitalism’s impulse to move, expand and swallow.’ It is therefore not enough to correct ‘biases’ or ‘abuses’ in order for AI to suddenly begin serving the interests of the countries of the South, since these biases and abuses are the very expression of the structural constraints which weigh on the development of AI, starting with the constraints of competition and the profit accumulation inherent to capitalism.

It is capitalism that is currently guiding and fuelling a global AI race which only benefits a handful of giant companies, most of them from the United States and China. As the United Nations development agency UNCTAD explained in a 2021 report: ‘At the country level, the United States is leading in AI development, with China rapidly catching up. These two countries accounted for as much as 94 per cent of all funding of AI start-ups between 2016 and 2020. The European Union is falling behind. Developing countries are at a disadvantaged position on AI development, particularly those in Africa and Latin America.’

The reason is quite simple – the more complex the AI systems you want to develop, the more you need: a) phenomenal computing power; b) astronomical quantities of data; and c) engineers and developer talents. These resources are today concentrated in the hands of Big Tech and their Chinese equivalents,³ which in turn take advantage of this to increase their lead and further widen the gap with the rest of the world.

Renewed extractivism and exploitation

In such a configuration, according to the influential Chinese entrepreneur and computer scientist Kai-Fu Lee: ‘The countries that are not in good shape are the countries that have perhaps a large population, but no AI, no technologies, no Google, no Tencent, no Baidu, no Alibaba, no Facebook, no Amazon. These people will basically be data points to countries whose software is dominant in their country. If a country in Africa uses largely Facebook and Google, they will be providing their data to help Facebook and Google make more money ….’⁴

More specifically, at present, countries in the South (with, of course, significant variations between them) tend to occupy the least enviable positions in AI value chains. First of all, we find them overwhelmingly in the role of supplier of raw materials and labour for the production of the material infrastructure of AI. Think the extraction of minerals such as lithium in Chile or cobalt in the Democratic Republic of Congo, as well as the immense assembly plants in China of the Foxconn company, which works as a subcontractor for the majority of the world’s largest IT companies.⁵ At the other end of the chain, there are also African and South-East Asian countries which today have the unfortunate distinction of ‘hosting’ most of the colossal quantities of digital waste that the world economy (rich countries in particular) generates each year.⁶

To these ‘classic’ forms of extractivism and exploitation are now added new forms that are specifically ‘digital’. This involves, in particular, the plundering of data extracted from these countries for free or almost-free, which fuels the development of high-value-added services that are then resold to them at a high price, within the framework of what UNCTAD describes as a form of ‘unequal exchange 2.0’. It is also about the millions of ‘click workers’ from the Global South who are paid a pittance to train algorithms or to root out offensive or illegal content from the Web, like the Kenyan workers paid $2 an hour by OpenAI to teach its well-known chatbot ChatGPT not to make racist or sexist comments.⁷

From neoliberal ‘laissez faire’ to the ‘Digital Cold War’

Naturally, many governments in the South are seeking to change this state of affairs, but the options for doing so are few. They can even prove counterproductive, in particular for states which think they can go it alone without calling into question the rules of the game. This at least is the opinion of Anita Gurumurthy, who writes: ‘[T]he desire to build local data infrastructures seems to go hand in hand with “AI partnerships” – a euphemism for easy access to citizen or public data by multinational firms with little or no overarching institutional norms. … Tech partnerships for public services delivery in developing countries thus come with huge risks. While they may bring efficiencies, they may well lead to a data exodus – transferring citizen data, often with very little privacy safeguards, to corporate AI labs.’

These risks are all the greater given that these same companies have been trying for several years to have international trade rules adopted which would further limit South countries’ room for manoeuvre in terms of digital sovereignty and industrialisation.⁸ Provisions tailor-made to defend the interests of Big Tech – such as ‘free flow of data across borders’ and ‘protection of source codes’ – have been included in a growing number of free trade agreements, such as the Trans-Pacific Partnership (since rebranded as the Comprehensive and Progressive Agreement for Trans-Pacific Partnership with the exit of the United States) and the United States-Mexico-Canada Agreement (USMCA).

Fortunately, in a surprising about-face, the United States has just announced that it will no longer support the inclusion of these contentious provisions in new agreements, starting with the agreement on electronic commerce that has been under discussion at the World Trade Organization (WTO) since 2019.⁹ But while the argument put forward – not to curb domestic regulation, particularly on subjects like AI – echoes the interests of the countries of the South, another, unstated reason is more problematic. We must also see in the US decision a desire to give itself the means to pursue a strategy of technological decoupling with Beijing, which would have been made more difficult with the adoption of a free trade agreement including China.¹⁰

There is a ‘Digital Cold War’ logic promoted by Washington which increasingly requires third countries – particularly from the South – to choose sides, making them de facto even more dependent on one of the two global digital superpowers. To avoid this trap, one step would consist of defending ‘digital non-alignment’ and the promotion of a global digital governance architecture decided within the framework of the United Nations.¹¹ But this is only the first step. As Anita Gurumurthy says, it will be difficult to imagine an AI that ‘work[s] for people and planet’ without a radical break with the current functioning of digital capitalism.¹²                             

Cédric Leterme is a researcher at Gresea (gresea.be), a small research and training centre based in Brussels dedicated to the promotion and diffusion of alternative discourses related to the functioning of the global economy, working closely with trade unions and NGOs but also with the wider public. The above article was originally published in French in Agir par la culture (www.agirparlaculture.be/).

Notes

1.   World Bank, ‘Harnessing artificial intelligence for development: A new policy and regulatory framework’, January 2020.

2.   La revue européenne des médias et du numérique defines ‘solutionism’ as a ‘current of thought originating in Silicon Valley which emphasises the capacity of new technologies to solve the world's major problems, such as disease, pollution, hunger or crime’.

3.   Amba Kak, Sarah Myers West and Meredith Whittaker, ‘Make no mistake – AI is owned by Big Tech’, MIT Technology Review, 5 December 2023.

4.   Cited in: Dave Gershgorn, ‘The list of countries that will benefit from the AI revolution could be exceedingly short’, Quartz, 26 March 2018.

5.   See: Sibo Chen, ‘“Immatérielle’’, l’expansion mondiale des TIC?’, Alternatives Sud, Vol. XXVII, No. 1, 2020.

6.   See, for example: Florence Lenoir, ‘De nos maisons aux décharges à ciel ouvert des pays du Sud Global, quel parcours pour nos déchets électriques et électroniques?’, Justice & Paix, 15 November 2021.

7.   Miguel Allo, ‘ChatGPT: des travailleurs kényans payés 2 $ de l’heure pour rendre le robot plus sûr, selon une enquête du Time’, RTBF, 19 January 2023. On ‘click work’ in general, see: Antonio Casilli, En attendant les robots – Enquête sur le travail du clic, Le Seuil, 2019.

8.   On this point, see: Cédric Leterme, ‘Bataille autour des données numériques’, Le Monde Diplomatique, November 2019.

9.   For a critical analysis of this decision, see: Cédric Leterme, ‘Volte-face des États-Unis sur le commerce électronique’, Le Vent Se Lève, 19 December 2023.

10. On this subject, see: Parminder Jeet Singh, ‘The U.S.’s signal of a huge digital shift’, The Hindu, 10 November 2023.

11. Parminder Jeet Singh, ‘Bras de fer États-Unis-Chine: nécessité d’un non-alignement numérique’, Alternatives Sud, Vol. XXVII, No. 1, 2020.

12. Anita Gurumurthy, ‘How to make AI work for people and planet’, openDemocracy, 10 March 2020.

*Third World Resurgence No. 359, 2024/2, pp 10-13