Rethinking AI Research from Global Majority Perspectives - Research Colloquium
- Edit Desk
- Feb 25
- 5 min read
By Joshua Baru, AI Researcher at Qhala
One of the three sutras of the recently concluded AI Impact Summit in India was Progress: advancing greater collaboration and knowledge-sharing among Global Majority actors. In this spirit, on 17 February 2026, more than 50 researchers convened for a Research Colloquium on the sidelines of the New Delhi Summit to ponder an urgent question: whose research will shape the intelligence age? Or, in the words of Didier Nkurikiyimfura, Senior Director at Smart Africa, in his keynote address at the Colloquium: “Will the Global Majority shape AI, or will AI shape us?”
AI Adoption?
Global AI companies and their partners have increasingly framed frontier/advanced AI adoption as the main route for emerging markets to realize AI’s promised benefits, via skilling, ‘AI for Good’ programs, and enterprise deployment partnerships. However, much as adoption of AI for the Global Majority is a worthy pursuit, researchers at the Colloquium posed an even nobler question: “How can we (finally!) dismantle the enduring legacies of colonialism that trap the Global Majority in vicious cycles of ‘technology adoption’ - without ever developing their own?”
This was well captured by Dr. Elizabeth Wamicha, a Social Impact Researcher at Qhala: “This Colloquium is intended to help researchers from the Global Majority move beyond the narrow procurement of foreign AI tools and instead co-design AI that reflects local needs.” This was also reiterated by Didier Nkurikiyimfura, who emphasized the necessity of co-creating AI tools as the Global Majority, rather than within country or continental silos: “We can build that future… if we work together. And the most important word here is together.”
For the Global Majority to sovereignly build AI systems, we must, as an indispensable prerequisite, invest in AI Research. This clarion call was captured by Prof Vukos Marivate during his keynote, chair of Data Science at University of Pretoria and a panelist at the recently announced Independent International Scientific Panel on AI. He observed that no African government meets the African Union’s benchmark of spending at least 1% of their GDP on R&D.
Yet, history keeps affirming why research precedes innovation: In 1970, Xerox Corporation made a long-horizon bet by establishing the Xerox Palo Alto Research Center (Xerox PARC) as a dedicated research lab for future information technologies. Out of that investment came breakthroughs that still power daily life; most notably Ethernet, and the practical, modern mouse.
If the Global Majority is to experience its PARC moment, then investment in AI research is indispensable.
AI Bubble?
There is much talk that we are in an AI Bubble. There is some truth to this, and we heed the counsel of some commentators that there is a likelihood that many AI companies are overvalued, and a correction in the markets is imminent.
That said, we push back on the label “AI bubble” and instead argue that what we are experiencing is more of an LLM bubble. We agree with Hugging Face CEO, Clem Delangue, who has argued that we’re in an “LLM bubble,” emphasizing that LLMs are a subset of AI and that broader AI applications can remain valuable even if LLM valuations reset. Researchers at the Colloquium were fully convinced that AI is here to stay - that we are not in an AI bubble - and that the promise of AI in making work easier holds true for the Global Majority.
Indeed, just as the enterprises that rule today’s corporate world are those that survived and thrived in the wild wild west of the internet era - the dot-com bubble - so we also argue that the organizations and countries that will lead Industry 4.0 and beyond are those that sovereignly develop and deploy AI systems that solve their people’s most pressing needs.
When the dot-com bubble burst (roughly 2000–2002), it did wipe out many internet startups and inflated valuations, but it did not stop the internet itself from expanding and evolving; instead, it helped usher in a more durable phase where surviving firms (and later entrants) built longer-term business models on top of the web. In the same way, we anticipate that once the Global Majority starts building AI beyond the current LLM-centered hype cycle, it will not be the end of AI, but its beginning for many within their borders.
It’s worth stressing that Artificial intelligence is not just large language models. AI is commonly defined as the ability of a computer to perform tasks associated with human intelligence, including reasoning, learning, and problem-solving. Beyond LLMs, AI includes (among other areas) machine learning, small language models, computer vision, speech recognition, recommender systems, anomaly and fraud detection, robotics and control, and reinforcement learning.
Introducing Minimum Viable Intelligence
Researchers at the Colloquium observed that the Global Majority is not a niche market; it is the demographic engine of the 21st century. Yet, a structural asymmetry exists: the region provides the humans, the data, and the raw labor, while the Global North captures the economic value of the Intelligence Age. The region represents 80% of the world’s population (thus the name Global Majority, moving away from the often derogatory ‘Global South’ label).
Despite the extreme diversity within Global Majority regions, researchers observed that they shared striking similarities: developer in Nairobi, a linguist in Santiago, and a researcher in Hanoi face the exact same constraints: exorbitant compute costs, non-representative datasets, and extractive governance models.
To overcome these constraints, Dr. Shikoh Gitau, CEO Qhala, proposed the Minimum Viable Intelligence model. She defined it as the smallest, context-aware layer of AI and human capability that reliably improves key local outcomes (like health, education, livelihoods or governance) at a cost, complexity, and risk level that low‑resource institutions can sustainably support.
She argued that as the Global North focuses on "Artificial General Intelligence" (AGI), the Global Majority must focus on "Developmental AI": Crop disease detection (Cassava in Africa), voice-first banking (UPI in India), and bureaucratic efficiency.
Researchers at the Colloquium agreed with this imploration, adding that sovereignty is not a slogan but a capability stack - built through talent, data governance, compute, and real deployment pathways - and they committed that the task now is to convert that stack into shared action. To ensure that the conversation continues and actual collaboration is borne out of the Colloquium, they formed working groups and committed to drafting shared principles, pooling infrastructure, and building trust mechanisms matters as a practical roadmap for South–South coordination.
Minimum Viable Intelligence is context-aware, usable, and sustainably improves local outcome. It is the hope of the convenors that this will not be remembered as just another convening, but as the moment the Global Majority began moving from paper to pavement.
We welcome you to read the full Research Colloquium report here. If you have any questions or wish to collaborate in developing South-South AI Frameworks as we advance Global Majority’s AI Ecosystem, feel free to contact us via research@qhala.com
Comments