From 11f042b35cb6265ffeee0a2d12c55e9f3ed9fe96 Mon Sep 17 00:00:00 2001 From: Diarra Yacouba Date: Wed, 26 Mar 2025 13:49:54 +0000 Subject: [PATCH] Edited introduction_to_maliba_ai.md (diarray) --- introduction_to_maliba_ai.md | 138 +++++++++++++---------------------- 1 file changed, 51 insertions(+), 87 deletions(-) diff --git a/introduction_to_maliba_ai.md b/introduction_to_maliba_ai.md index 14713d7..cb3e320 100644 --- a/introduction_to_maliba_ai.md +++ b/introduction_to_maliba_ai.md @@ -1,131 +1,95 @@ -# An Introduction to MALIBA-AI +## An Introduction to MALIBA-AI -## Table of Contents +### Table of Contents -- [The MALIBA-AI Initiative](#i.-the-maliba-ai-initiative) -- [Philosophy](#iii.-philosophy) -- [Objective & Plans](#v.-objective-and-plans) -- [Fields of Interest](#vi.-field-of-interest) -- [Who is MALIBA-AI for](#vii.-who-is-maliba-ai-for) -- [Publications](#viii.-publications) - -- [Copyright & License](#x.-copyright-&-license) -- [Conclusion](#xi.-conclusion) - +- [The MALIBA-AI Initiative](#i-the-maliba-ai-initiative) +- [Philosophy](#ii-philosophy) +- [Objective & Plans](#iii-objective--plans) +- [Field of Interest](#iv-field-of-interest) +- [Who is MALIBA-AI for](#v-who-is-maliba-ai-for) +- [Publications](#vi-publications) +- [Copyright & License](#vii-copyright--license) +- [Conclusion](#viii-conclusion) ## I. The MALIBA-AI Initiative -        MALIBA-AI is defined as a community-driven AI initiative with a name derived from Mali Ba (Great Mali) and AI (Artificial Intelligence), pronounced as Mali-Ba-AI. It is an initiative set in motion with a simple but strong objective: solve problems using artificial intelligence in Mali while ensuring no Malian is left behind by technological advances. - -        Over the last half a century there has been an explosion in the field of Artificial Intelligence in general, from the early development of expert systems and neural networks at research institutions like MIT and Stanford, to the emergence of machine learning algorithms, to the advances in natural language processing, all the way to the works presently being done by many AI-focused companies such as OpenAI, Anthropic, DeepMind, and more. - -        Nothing really has impacted our day to day life in mass numbers like this prior to the introduction of artificial intelligence systems to my recollection. The integration of AI has proven fruitful in our lives. Note that I will emphasize on AI models here, computing infrastructure essentially is a medium in which AI is processed, and delivered, although it is an extremely important part of an AI system. I like to think that the most valuable part is the AI model itself, I am sure a hardware engineer working on specialized AI chips would disagree. This is a philosophical discussion I will have to get into in a later post. For now, let us just assume that the AI model is the most valuable part. +MALIBA-AI is a community-driven AI initiative. The name comes from "Mali Ba" (Great Mali) and AI (Artificial Intelligence), pronounced as Mali-Ba-AI. Our goal is simple yet powerful: use artificial intelligence to solve local problems while ensuring no Malian is left behind. -        With that assumption, AI's power in general has been fairly impactful in the developed world, everything has been enhanced by artificial intelligence well almost everything. AI-powered automation has transformed industries in some form or shape. Manufacturing, the financial market, military, healthcare, transportation all of these use machine learning models to perform as they do these days. This widespread adoption of AI technology, although to some extent has reached the non-developed countries, it has not been as rewarding as it has in the developed world. +Over the last half-century, the field of AI has exploded—from expert systems and neural networks at institutions like MIT and Stanford, to the rise of machine learning and natural language processing, and now to cutting-edge companies like OpenAI, Anthropic, and DeepMind. -        A country like Mali is missing out on all of the advantages that comes with AI adoption. A prime example of this is the health care system, food production, financial market, educational system. All in all the entire nation's AI infrastructure is underdeveloped, or completely non-existent. There is no denial that AI-powered platforms, like Meta and Google have reached a number of people. A 2018 report1 from the International Telecommunication Union(ITU) reported that 12.7% of the nation are internet users, with a 3.2% households that own a computer. These figures are for a country with a growing population of 23.77 million as yielded by a Google Search. These are some very depressing numbers and not to mention that vast number of those internet users are social media users(not in the occasional use sense), these are the typical users who only go online for the mere use of a social media platform. I do not have an exact figure on Meta users, but based on personal experience the platform has a really strong user base. Which I must point out is totally fine. +Nothing in recent history has impacted our daily lives at this scale. AI integration is transforming the world. While computing infrastructure is essential, we focus on the AI model itself as the most transformative component—though that’s a debate for another post. -        However, beneath these statistics lies a more fundamental barrier – language. Most Malians do not speak French or English, the languages that dominate today's AI interfaces. The majority of Mali's population speaks indigenous languages like Bambara, Fulfulde, and Songhay. These languages are primarily oral rather than written, which creates an additional barrier to technology access when most interfaces require text input. +With that assumption, AI's power has been impactful in the developed world—everything has been enhanced by artificial intelligence, well, almost everything. AI-powered automation has transformed industries in some way. Manufacturing, finance, healthcare, transportation, and education are now deeply influenced by AI. But in many developing countries, including Mali, these advancements have not been equally rewarding. -        This linguistic divide creates a profound inequality in technological opportunity. Consider education – children learn most effectively in their mother tongue, yet educational technology requires knowledge of colonial languages. Consider healthcare – AI diagnostic tools could revolutionize rural healthcare, but not if doctors and patients cannot interact with them in their native language. Consider agriculture – AI could help Malian farmers optimize their practices, but not if the farmers cannot understand the interface. +This is especially true for sectors such as healthcare, agriculture, finance, and education. Overall, the entire country's AI infrastructures are underdeveloped or non-existent. Despite the reach of AI-powered platforms like Meta and Google, their impact remains surface-level. A 2018 report from the International Telecommunication Union (ITU) showed that only 12.7% of Malians used the internet, with just 3.2% of households owning a computer [1]. These figures are for a country with a population of around 23.77 million in 2023, according to the world bank [2]. These are sobering numbers—and most of these internet users primarily access social media. We can roughly estimate that while they command a significant portion of web traffic, social media platforms constitute a tiny fraction of total websites—likely well below 0.5%. Thus, social media-only users typically miss out on most internet resources. -        The language barrier is not just an inconvenience – it is a fundamental obstacle to Mali's participation in the AI revolution. When technology requires users to speak French or English, we create a two-tier society where only the formally educated can participate in technological advancement. The transformative potential of AI – the democratization of specialized knowledge, the automation of routine tasks, the unlocking of human potential – remains inaccessible to most Malians. +Yet beneath the digital divide lies an even deeper barrier: language. Most Malians speak indigenous languages like Bambara, Fulfulde, and Songhay—not French or English, the languages that dominate AI systems and interfaces. Since these languages are primarily oral, the requirement for text input becomes a major obstacle. -        This exclusion is not inevitable. AI itself offers the solution to this problem. With advancements in natural language processing and speech recognition, we now have the capability to create AI systems that understand and speak Malian languages. We can build interfaces that work through natural conversation rather than text, making technology accessible to everyone regardless of literacy or formal education. We can ensure that the AI revolution leaves no Malian behind. +This linguistic divide creates real inequality. Consider education: children learn more effectively in their mother tongue, yet top educational technologies require knowledge of colonial languages. In healthcare, diagnostic AI tools can't help if patients and providers can’t interact in their language. In agriculture, farmers can’t use AI-driven platforms they don’t understand. -        The insight that sparked MALIBA-AI came from a simple observation: "It is a plus to know how to code, but it will not be a requirement anymore." As AI advances, people who don't know programming can still create software, websites, and digital solutions through natural language interfaces. But this democratization of technology creation only benefits those who speak the languages AI systems understand. +The transformative potential of AI—the democratization of specialized knowledge, the automation of routine tasks, the realization of human potential—remains inaccessible to most Malians. -        MALIBA-AI is here to challenge this linguistic exclusion. Imagine a future where a farmer in Sikasso who speaks only Bambara can build a website to sell his produce by simply having a conversation with an AI in his native language. Imagine a mother in Timbuktu who can access healthcare information for her child by speaking to an AI assistant in Songhay. Imagine students across Mali learning physics, mathematics, and literature in their mother tongues through AI-powered educational platforms. +But this exclusion isn’t inevitable. In fact, AI itself offers the solution. With advances in natural language processing and speech recognition, we can build systems that speak and understand Malian languages. We can create speech-first interfaces that don’t require literacy or formal education. We can build a future where AI includes everyone. -        This is not a distant dream but an achievable reality with today's technology. MALIBA-AI is focused on developing speech-first AI systems that understand and respond in Mali's languages, making technology accessible to all Malians regardless of formal education or literacy. Our community-driven approach recognizes that international AI companies may not prioritize Mali's linguistic needs for years, if ever – so we must build these solutions ourselves. +The spark that lit MALIBA-AI came from this observation: "It’s a plus to know how to code, but it won’t be a requirement anymore." With AI, even people with no programming background can build apps or digital tools by talking to a machine. But this only works if the AI speaks your language. -        The impact of bringing AI capabilities to all Malians, not just the formally educated, would be transformative: +MALIBA-AI aims to challenge this exclusion. Imagine a farmer in Sikasso building a website by talking to an AI in Bambara. A mother in Timbuktu getting medical advice in Songhay. Students learning science through AI-powered platforms in their native tongue. -        **In Education**: Students learn most effectively in their mother tongue. AI tutors that speak Malian languages could provide personalized education to every child, helping them grasp concepts more quickly and deeply than in a colonial language. +This is possible today. Our mission is to build speech-first AI systems for Malian languages so that technology becomes accessible to all. International companies may overlook us—but we can build it ourselves. -        **In Agriculture**: Farmers could access weather forecasts, crop disease identification, market prices, and optimization techniques through conversation in their own language, increasing productivity and income. +**Impact Areas:** -        **In Healthcare**: Patients could describe symptoms in their native language to AI diagnostic assistants, helping healthcare workers provide better care with limited resources. +- **Education**: AI tutors in local languages could offer personalized support to every child. +- **Agriculture**: Voice interfaces could deliver crop advice, weather alerts, and market data in Bambara. +- **Healthcare**: Patients could explain symptoms and receive help in their native language. +- **Business**: Entrepreneurs could use AI tools to manage businesses without needing to learn French or English. -        **In Business**: Entrepreneurs without formal education could build websites, manage inventory, create marketing materials, and access global markets through AI tools that understand them. - -        By removing the language barrier, we remove the most fundamental obstacle to Mali's participation in the AI revolution. We enable all Malians to benefit from AI's power to democratize knowledge, automate routine tasks, and unlock human potential. This is the vision that drives MALIBA-AI – a future where language is no longer a barrier to technological empowerment. +By removing the language barrier, we remove the most fundamental obstacle to Mali's participation in the AI revolution. ## II. Philosophy -        MALIBA-AI follows the same model as most open-source projects. MALIBA-AI aims to take a community-driven R&D approach in its works. Any of the projects published under the cover of MALIBA-AI will not be for-profit. All of the projects are concentrated for now, on a single geographical area and that is Mali, with special attention to its linguistic diversity. The initiative aims to build models that will focus on a specific problem in Mali, and one problem at a time. Although I occasionally think of it as an open-research, the reality is that it is a collaborative community research, driven by enthusiasts, engineers, linguists, researchers, and students passionate about leveraging AI to address Malian challenges. This initiative has come to life as a passion project, hence it ought to be worked on with that mindset. A passionate collective looking to make an actual change with their skillsets, ensuring that no Malian is left behind by technological advances. - - -## III. Objective and Plans +MALIBA-AI follows a community-driven, open-source model. It’s not-for-profit and focused entirely on Mali’s specific challenges, especially its linguistic diversity. We work on one problem at a time, developing models and tools to tackle real-world issues. -        The intent is to have at least a documentation on all of the aforementioned projects, and those not mentioned yet, describing if they will work, or if they will not for the country and if they do how to get it up and running as quickly as possible. All the projects are regarded as community research projects, going through trial and error, hypotheses, and experiments. Because of the complexity of the subjects and the convergence of multiple fields projects are worked on in a collaborative manner, there is no specific development timeline. It should be clear that these projects will not be an overnight work. The primary plan however for the time being is to introduce the concept, build a strong community of contributors, and work on projects from our project stack, and see how it goes from there. +We see this as collaborative community research, not traditional open research. It’s driven by engineers, linguists, researchers, and students—people passionate about using AI for good in Mali or maybe just passionate about our Mali-Ba. -## IV. Field of Interest - -        MALIBA-AI fields of interest are focused on language technologies. To elaborate they are the following: Language Models, Speech Technologies, and Accessibility Software Solutions. These are the accumulation of the fields that I believe could make a major impact in the current Malian climate. Let's take look at each in an introductory manner. - -### Language Models +## III. Objective & Plans -        This is a cornerstone of modern AI. It is the foundation upon which we build solutions that understand, generate, and translate language. MALIBA-AI intends to develop foundation models specifically trained for Malian languages, creating neural machine translation systems connecting Mali's diverse linguistic landscape. Most importantly, these models will focus on the primarily oral nature of Malian languages, enabling people who cannot read or write to still interact with technology through speech. Learning happens most naturally in one's mother tongue - these models will ensure that technology becomes accessible regardless of the language a person speaks, enabling education, business development, and innovation from previously excluded communities. +Our goal is to document and test each idea: will it work for Mali? If so, how can we build and deploy it effectively? Projects go through trial and error, backed by community-driven experiments. -### Speech Technologies +We don’t expect overnight success. For now, the plan is to grow the community, introduce the concept, and begin work on practical solutions from our project stack. -        Speech Technologies are perhaps the most crucial component of our work. In a society where languages are primarily spoken rather than written, and where many people have not had formal education, speech interfaces are the gateway to technological inclusion. Speech recognition and synthesis systems provide an interface that is natural and accessible for all Malians, regardless of literacy level or formal education. - -        This field is essential for democratizing access to technology in Mali, where many citizens who have been excluded from previous technological waves could now participate fully through natural spoken conversation. MALIBA-AI's speech technologies will enable villagers, elders, market traders, farmers, and anyone without literacy in colonial languages to create content, access information, and harness AI tools through voice interaction in their native language. This represents true technological empowerment and has the potential to unleash previously untapped human potential across Mali. - -### Accessibility Solutions +## IV. Field of Interest -        Although AI commonly implies sophisticated systems, MALIBA-AI is taking a different approach to AI accessibility, focusing on solutions in Malian context. MALIBA-AI intends to develop solution on top of its developed AI models that can address local problems, and interfaces designed for users with varying levels of technical literacy. I should mention that by accessibility I am referring to making technology truly accessible to all Malians, regardless of educational background, linguistic preference, or technological resources. +MALIBA-AI currently focuses on one core area: **speech systems for Malian languages**. Our aim is to build technologies that make it possible for people to interact with AI using voice in their native language—without needing to read or write. -Community empowerment is discussed further in MALIBA-AI's whitepaper. +We believe that this approach will be the key to making technology accessible to Malians of all educational backgrounds. ## V. Who is MALIBA-AI for -        MALIBA-AI is for everyone. We welcome all who share our vision of linguistic inclusion through technology. Our community unites AI practitioners, software engineers, language experts, educators, artists, community members, and technology enthusiasts. Every skill contributes to our mission, whether you're developing models, preserving languages, testing applications, or raising awareness. Together, we're creating an inclusive technological ecosystem where no Malian is left behind. +MALIBA-AI is for everyone who believes in inclusive technology. Our community includes AI practitioners, developers, educators, linguists, artists, and everyday citizens. Whether you help build models, test tools, raise awareness, or preserve languages—there’s a role for you. ## VI. Publications -        The type of publications that will be published by MALIBA-AI are model cards, software documentation, blog posts, whitepapers, and academic research papers. - -### Blog - -        Blog posts as mentioned in MALIBA-Blog section are occasional publications for updates on projects and community activities. - -### Whitepaper - -        Serve as the full specifications for a project, this include hypothesis, modeling, and necessary information for the project. Except for the main web app, every project will have an accompanying whitepaper. - -### Research Paper - -        A fully fledged research paper is published on every project as evidence, following open science principles, and for those interested in having an in-depth understanding of the project's methodologies and findings. - -Note that at the moment we are in no way expert scholars, we are a community of enthusiasts working on problems we believe we might be able to do something about. - -### Model Documentation - -        It is impossible to have a successful open source AI project without documentation. All AI model related projects will have a full documentation for use, training, evaluation, and contribution. - -## VII. Versioning & Future - -        No one really knows what the future holds, all we can do is plan as precisely as we can. The only future plan of MALIBA-AI is to grow our community, expand our language coverage to all Malian languages, develop increasingly solutions tailored to local needs, and ensure that our solutions remain free, open, and accessible to all Malians. It is after all a community-driven initiative independent of anything that can really hold it back. - -## IIX. Copyright & License +MALIBA-AI will publish the following: -        MALIBA-AI will make attempt to avoid proprietary items for all of its projects. All the projects will try to build everything from scratch or otherwise use an open-source that has a flexible license. Any of the projects that we work on will be open-source, as we are more interested in implementing solutions than the credit/financial gain aspect of things. Due to our commitment to open access and community ownership, all projects will be hosted in public repositories with permissive licenses that allow for wide use, adaptation, and contribution while ensuring that derivatives remain open to the community. +- **Blog Posts**: For updates and community highlights. +- **Whitepapers**: These are in-depth project specs—like research proposals—that cover goals, hypotheses, and methodologies. +- **Research Papers**: For peer-reviewed or formal documentation of experiments and findings. +- **Model Documentation**: Open guides for usage, training, and contributions to our AI models. -## IX. Conclusion +*Note: We are not academic scholars—we're an enthusiast community tackling real problems that we think we can do something about.* -        We've reached the conclusion of this introductory post. MALIBA-AI is a community-driven initiative to solve the most pressing problem facing Mali in the AI revolution: the linguistic exclusion of the majority of our citizens. When technology requires users to speak French or English, we create a two-tier society where only the formally educated can participate in technological advancement. +## VII. Copyright & License -        Using Language Models, Speech Technologies, and Accessibility Solutions, we aim to create a future where a farmer in a remote village can build a website through conversation in Bambara, where a student can conduct research in Songhay, where entrepreneurs can harness AI without first mastering a foreign language. The social and economic impact of such technological inclusion would be transformative for Mali. +All projects will be open-source. We avoid proprietary dependencies wherever possible, and everything will be hosted with permissive licenses. Our goal is wide accessibility, reusability, and community ownership. -        International AI companies may not prioritize Mali's languages for years, if ever, they see our market as too small. This is why we must build these solutions ourselves, for ourselves. MALIBA-AI's goal is ambitious, but even achieving a fraction of this vision could democratize access to technology, unleash previously untapped human potential, and fundamentally change Mali's technological trajectory. When all Malians can participate in the AI revolution, regardless of the language they speak, we will have truly ensured that no Malian is left behind. +## VIII. Conclusion - +[1] https://www.itu.int/net4/itu-d/icteye/CountryProfileReport.aspx?countryID=150 +[2] https://datacatalogfiles.worldbank.org/ddh-published/0038126/DR0046429/POP.pdf