Artificial intelligence, nurtured within the American private sector, is emerging as a novel and potentially troubling force, reshaping our understanding of empire. Traditional decolonization, rooted in the 18th-century Atlantic revolutions, aimed to dismantle European imperialism. The contemporary decolonization movement, epitomized by Black Lives Matter, mobilizes a new generation of thinkers and activists.
Decolonization involves confronting the factors that facilitated European colonization, such as exploitative business models and the violent imposition of social hierarchies. It also necessitates recognizing the enduring repercussions of colonization, including present-day economic and environmental injustices and the loss of intellectual and cultural equity. Until a comprehensive decolonization is achieved, these same factors will continue to shape the development and deployment of AI, rendering AI a vector of a different form of colonial harm.
However, just as AI has the potential to reshape our notions of empire, it could also facilitate new alliances that promote a fresh concept of freedom.
One prominent domain where AI holds significant promise for humanity is healthcare. AI can enhance the speed, precision, and innovation in healthcare systems, from diagnostics to treatment delivery. Nevertheless, healthcare systems are fundamentally compromised by the historical erasure of non-European communities, which has profound implications for contemporary health AI.
AI models rely on data for training, and the quantity and quality of relevant data significantly influence an AI model's accuracy and utility. A case in point is the use of AI in genomic research.
Every living organism possesses its unique set of genes. Genomic data richness enables scientists to discern generational patterns and genetic variations that may trigger various effects, aiding in drug development and understanding the genetic basis of diseases in humans. AI excels at detecting deep patterns in vast datasets, making it well-suited for the complexity of genomic data analysis.
However, a Nature Medicine report in 2022 estimated that a staggering 86% of global genomic research is conducted on individuals of white European ancestry, who represent only about 12% of the world's population. Consequently, AI models derived from such data may not be effective for the remaining nearly eight billion people worldwide.
Data Gaps
The absence of genomic data from underrepresented communities can hold crucial insights for developing new medications and treatments benefiting everyone. Some genetic variations may exist exclusively in certain populations. While AI could expedite drug discovery, these gains are overshadowed by data gaps that hinder comprehensive research. Building AI models on incomplete or non-representative data exacerbates research disparities.
Decolonizing AI is imperative for realizing its potential for public good in various domains. The historical erasure of communities also extends to contemporary national statistics, presenting analogous challenges for AI development in the public sector.
Though AI implementation could enhance government efficiency, structural issues in population data undermine its effectiveness. This affects government functions across education, justice, transportation, energy infrastructure, housing, and employment policy, incurring costs and missed opportunities for society.
While AI is often lauded for its societal benefits, research conducted in 2022 by Abeba Birhane indicated that only about 15% of the most influential AI research papers addressed societal needs, with a mere 1% considering potential negative aspects. Much of AI's evolution mirrors the extractive business models that fueled European colonization, perpetuating similar harmful consequences. In 2021, 'stochastic parrots' research by Timnit Gebru warned that the marginal performance gains in generative AI were outweighed by soaring operational costs and environmental impacts, disproportionately affecting marginalized communities globally.
Recent investigations have exposed the exploitative nature of 'ghost work,' where low-income workers in developing countries are paid meager wages, sometimes as low as $10 per day, to label datasets for corporate AI systems. This form of labor is now even being offered to prison inmates in Finland for less than $2 per hour.
European colonization was not solely about extraction; it also involved suppressing and controlling colonized populations to perpetuate extraction. Investigative work in 2022, led by Karen Hao for the MIT Technology Review, unveiled how AI tools were being used to exploit the global poor, create digital apartheid, and foster political disenfranchisement and coercive surveillance worldwide.
In traditional sectors, Environmental, Social, and Corporate Governance (ESG) reporting has been introduced to mitigate colonial legacies in global supply chains and business activities. However, the absence of equivalent ESG reporting in the digital realm hampers progress toward sustainable development goals shared by all.
AI for Inclusive Flourishing
This is a global predicament that will worsen as traditional organizations expand their digital capabilities and the global digital economy burgeons. If the narrow imagination imposed by European colonization facilitated colonization, then decolonization entails fostering a more diverse range of perspectives.
Thus, the movement to decolonize AI includes initiatives such as researchers at DeepMind advocating for the de-centering of Western norms, communities at MozFest creating decolonial alternatives to voice and facial recognition technology, and the Indigenous Protocol and AI Working Group exploring indigenous epistemologies and ontologies to contribute to AI development for the benefit of all living beings.
Decolonizing AI transcends merely rectifying past harms; it involves learning from history's mistakes to ensure AI contributes to a better future for all. As stated in the AI Decolonial Manyfesto, "Our destinies are intertwined. We owe each other our mutual futures."
European colonization thrived on competition among imperial powers and geographical separation from the consequences of colonization. Today, thanks to decolonization, we live in an era of international cooperation and diverse perspectives that enrich every facet of our lives. These gains should also underpin our aspirations for AI.
Freedom is more than independence; it involves flourishing through mutual, respectful interdependence with one another and the world. AI stands poised to become a pervasive technology, transcending sectors, domains, and borders with unprecedented speed and power. None of us can afford to make mistakes; decolonizing AI can guide us toward the right path.

إرسال تعليق