Listen to this Post
Artificial intelligence (AI) has quickly become a fundamental part of our daily lives, especially since the rise of tools like OpenAI’s ChatGPT in 2022. However, despite being promoted as global solutions for all, AI products have predominantly catered to the interests of Western nations, leaving out many regions, including Africa. In response, a group of African researchers and technologists are challenging this imbalance by creating AI solutions tailored to their local needs, addressing issues that the traditional Western-dominated tech industry has ignored.
AI has undoubtedly revolutionized sectors from healthcare to education, yet its development and deployment have been shaped largely by the interests of the United States and Europe. The narrative of AI being a “global” tool is often contradicted by its heavy focus on Western languages, cultures, and socioeconomic structures. For many African countries, this leaves them without representation in the technologies that are shaping the world.
AI’s Global Power Imbalance and the African Response
The global AI industry, controlled largely by major tech companies based in the West, has often ignored the needs and realities of regions like Africa. While AI systems promise to democratize access to technology, they often overlook the social, economic, and cultural contexts of non-Western regions. A group of African researchers, particularly those affiliated with the Distributed AI Research Institute (DAIR), is working to change this status quo by focusing on AI solutions that prioritize the needs of historically marginalized communities in Africa.
One of the prominent figures in this movement is Nyalleng Moorosi, a senior researcher at DAIR. Moorosi’s work is deeply influenced by her background in machine learning and her experiences as an educator in South Africa. Through her involvement in the Deep Learning Indaba, an organization aimed at strengthening AI in Africa, Moorosi emphasizes the importance of building AI systems that cater to African needs rather than reinforcing Western-centric models. She argues that AI should not just be for the rich and powerful, but should serve communities that have long been excluded from technological advancements.
In 2018, Moorosi, along with DAIR’s founder Timnit Gebru and fellow researcher Raesetje Sefala, began a project that examined the impact of apartheid on the development of South African townships. By analyzing satellite images, they created a dataset to track the changes in these neighborhoods over time. However, when they sought to publish their findings, many Western academic institutions dismissed their work, arguing that it was not “true” AI research. This response highlighted the biases embedded within the Western-dominated academic and tech sectors, where the value of AI research is often determined by its relevance to Western contexts.
Cultural and Linguistic Gaps in AI
One of the most significant challenges facing AI development in Africa is the lack of representation of African languages and cultures in major AI systems. Asmelash Teka Hadgu, co-founder and CTO of Lesan AI, has worked extensively on building AI models that address the linguistic diversity of Africa. Lesan AI focuses on low-resource African languages, such as Amharic and Tigrinya, which are often ignored by larger AI models from companies like OpenAI.
Hadgu’s work highlights a fundamental issue: AI models developed by Western companies are often ill-equipped to handle African languages, which are not as widely represented on the internet as English or European languages. In fact, research has shown that African languages are among the least supported in AI models. OpenAI’s ChatGPT, for example, performs poorly in languages like Amharic and Tigrinya, often producing nonsensical outputs. Hadgu argues that these shortcomings reflect the bias built into Western AI models, which prioritize English and other widely spoken languages while ignoring the needs of millions of African speakers.
Additionally, the data used to train these models is often skewed toward the Western world. A study by the Data Provenance Initiative revealed that 90% of the data used to build AI models comes from Europe and North America, with only 4% from Africa. This lack of representation not only perpetuates linguistic and cultural gaps but also reinforces existing inequalities in access to technology.
The Challenge of Data Sovereignty
Another issue facing African AI researchers is the exploitation of local knowledge by big tech companies. While smaller AI startups in Africa, like Ghana NLP, work to develop models tailored to local languages and cultures, they often face challenges in retaining control over their work. Large tech companies frequently “open-source” these models, allowing them to use local data and ideas without compensating the creators. This situation, known as “data sovereignty,” raises concerns about the exploitation of African knowledge without adequate benefit for local communities.
Dr. Paul Azunre, co-founder of Ghana NLP, has firsthand experience with this issue. He recalls how Facebook used the data from their models without compensating them, highlighting the power imbalance between small African startups and tech giants. For Azunre, the priority is ensuring that local communities are at the forefront of AI development, with control over their data and the benefits that come from it.
What Undercode Says:
The struggles faced by African researchers and technologists in the AI space reflect a larger global issue: the dominance of Western tech companies in shaping the future of technology. The African continent, despite being home to over a billion people, remains largely excluded from the decision-making processes that determine how AI technologies are developed and deployed. However, the work of organizations like DAIR and individuals like Moorosi, Hadgu, and Azunre offers a promising glimpse into how AI can be adapted to meet the unique needs of Africa.
One of the most striking aspects of their work is the emphasis on community-centered AI solutions. Unlike the profit-driven motives of large tech companies, these researchers aim to create systems that serve the people who have historically been left behind by technological progress. Whether it’s using satellite imagery to track changes in South African townships or developing AI models for low-resource African languages, the goal is to ensure that AI becomes a tool for empowerment, not exploitation.
The broader implications of this work challenge the current AI industry to rethink its priorities. While Western companies continue to focus on building AI systems that cater to wealthy, English-speaking users, African researchers are pushing for a more inclusive approach—one that recognizes the diversity of human experience and ensures that AI serves all people, not just the privileged few. As AI continues to evolve, the voices of researchers from Africa and other marginalized regions will be crucial in shaping a more equitable and inclusive future for technology.
Fact Checker Results:
- The claim that AI systems are primarily designed for Western audiences is supported by evidence from studies on the language capabilities of major AI models, which often fail to support African languages.
- The issue of data sovereignty is a real concern, with multiple instances of African startups being exploited by larger tech companies without compensation for their work.
- The trend of African researchers focusing on community-centered AI solutions represents a significant shift in the development of AI technologies, as these solutions aim to address the specific needs of African societies.
References:
Reported By: www.zdnet.com
Extra Source Hub:
https://www.medium.com
Wikipedia
Undercode AI
Image Source:
Unsplash
Undercode AI DI v2
