On 23 January 2024, Microsoft Unlocked published Advancing AI responsibly<\/a><\/em> \u2014 the public output of the 2023 Stimson Center \/ Microsoft Office of Responsible AI Global Perspectives on AI Fellowship. The piece was introduced by Natasha Crampton, Microsoft’s Chief Responsible AI Officer, and concluded by Brian Finlay, President and CEO of the Stimson Center. Between the introduction and the conclusion, fellows from fourteen countries \u2014 including Aziz Soltobaev for Kyrgyzstan \u2014 set out what they had been arguing inside the cohort over the previous year.<\/p>\n\n\n\n The Fellowship itself was a year-long programme run jointly by the Henry L. Stimson Center in Washington, DC, and Microsoft’s Office of Responsible AI. It brought together practitioners from civil society, academia, and non-profit organizations across fourteen countries \u2014 Kyrgyzstan, Egypt, Nigeria, Sri Lanka, Turkey, Kenya, and Latin American and Central European partners among them \u2014 to discuss responsible AI from inside the contexts where AI systems land rather than the contexts where they are designed. Aziz Soltobaev took part as the Kyrgyzstan fellow, presenting on Topic 1: Importing AI<\/em>.<\/p>\n\n\n\n The Kyrgyzstan contribution drew on two examples \u2014 both about what happens when technology designed elsewhere is imported into a context that was not part of the original design conversation. The first was Estonia’s X-Road<\/strong> data exchange architecture. Estonia built X-Road for its own state and population, but the design principles travelled: Tunduk in Kyrgyzstan, e-government interoperability work elsewhere in the region. Importing the architecture worked in part because Estonia had built the surrounding institutional context openly enough to be studied, adapted, and reshaped by countries with different conditions.<\/p>\n\n\n\n The second example was Google Translate’s Kyrgyz support<\/strong>. The platform’s initial coverage of Kyrgyz was uneven \u2014 predictable for a language with limited training data. What changed it was not a corporate initiative; it was community contributions, where Kyrgyz speakers steadily corrected and extended the model’s outputs from inside Bishkek and the regions. Imported AI worked, in this case, because the import was not finished on arrival \u2014 local users completed it.<\/p>\n\n\n\n The framing for both: cultural context and respect for local values are not soft notes appended to a deployment. They are the difference between an imported system that functions in Bishkek and one that does not.<\/p>\n\n\n\n The Fellowship’s underlying message \u2014 and the message that comes out of the Unlocked publication \u2014 is a directional one. For Global South countries on the receiving end of AI supply chains, useful precedent does not only sit North and West, where the systems are designed and the standards are set. It also sits South and East, among countries doing the same adapting work \u2014 often in parallel, often without a forum to compare notes. The instruction is plain: look forward, look around, and look for the best cases broadly. The Fellowship was one of the rare structured spaces in which the looking-around could happen.<\/p>\n\n\n\n \u2192 Read the publication: Microsoft Unlocked \u2014 Advancing AI responsibly<\/strong><\/a><\/p>\n\n\n\n \u2192 Background on the fellowship cohort: Importing AI: Notes from the Stimson Center Global Perspectives on AI Fellowship (2023)<\/strong><\/a><\/p>\n\n\n\n
\n\n\n\n