AI isn’t equal for everyone and Microsoft’s Project Gecko wants to fix that

Microsoft has unveiled an innovative initiative known as Project Gecko, designed to create affordable and customizable AI systems tailored for communities often overlooked in traditional AI training datasets. The tech giant highlights that, despite the advancements of generative AI in various applications, these technologies frequently underperform in numerous languages and fail to capture the diverse social and cultural contexts of different populations. Project Gecko seeks to tackle these issues by prioritizing equitable and inclusive AI development. Ashley Llorens, Corporate Vice President and Managing Director of Microsoft Research Accelerator, emphasized in a recent blog post that, "Building AI systems from the ground up, shaped by the knowledge, languages, and modalities of the global majority, yields more innovative, useful solutions for a great number of people. This marks a critical advancement in our efforts to adapt and deploy AI on a broader scale in low-resource environments." The initiative focuses on creating AI systems that incorporate local languages, culturally relevant content, and multimodal interactions through text, voice, and video. To achieve this, Microsoft has assembled a team of researchers from various locations, including Microsoft Research Africa in Nairobi, Microsoft Research India, and the Microsoft Research Accelerator in the United States, in collaboration with Digital Green. As part of Project Gecko, Microsoft has also developed a new AI framework called MMCTAgent (Multimodal Critical Thinking Agent Framework). This framework is designed to analyze inputs from speech, images, and videos, providing context-aware and relevant responses. Microsoft asserts that this system aims to enhance experimental frontier models by supporting specialized tools that broaden their functionality. Recognizing that many affordable connected devices in the Global South lack the processing power for large language models (LLMs), the Project Gecko team has opted to work with Small Language Models (SLMs) instead. The initiative's initial focus targets small farms in India and Kenya, with plans to extend its applications into sectors like healthcare, education, and retail. Moreover, Microsoft has reported that Project Gecko technologies have improved FarmerChat, an AI assistant by Digital Green that operates primarily through voice. This technology enables farmers to submit queries via speech or text, receiving actionable responses complete with step-by-step instructions in their preferred language, delivered through text, voice, and relevant video content. Field studies conducted in India and Kenya have demonstrated enhanced response quality, usability, and user trust in comparison to existing state-of-the-art models.