Machine reasoning (AI) through Natural Language commands (Generative AI) for mission projects in remote fields (off-line)
Rodrigo Tinoco
January 2025
automatic machine translation
Portuguese to English
x. Description of the problem
In the global missions landscape, we have seen a significant shift in the origin of the missionary force, which has migrated from the Global South to the so-called "majority world". However, despite this transition, innovation processes, financial resources and other essential tools have not yet kept pace with this change, leaving many mission organizations in remote regions at a disadvantage.
These organizations play a crucial role on several fronts, such as humanitarian assistance, church planting, Bible translations and other vital initiatives. However, they face serious technological limitations that compromise their effectiveness and impact. The lack of access to modern tools, especially in areas without an internet connection, represents a significant obstacle to the advancement of their missions.
This project emerges as a solution to democratize access to Generative Artificial Intelligence (Generative AI), offering tools that work directly on local computers, without the need for an internet connection. In this way, we aim to empower these organizations by expanding their capacity for communication, content creation and information management, even in remote contexts and with limited resources.
x. Clarification note: Generative AI
In today's global missions landscape, Generative Artificial Intelligence (Generative AI) has been widely discussed, especially on social media. However, many of these discussions lack depth and clarity, leading to misunderstandings about the potential and limits of this technology. In the Christian and missionary context, this lack of understanding often results in resistance to the use of AI, viewed with suspicion or skepticism.
The reality is that Generative AI should not be seen as a threat or substitute for human intelligence, but as a powerful tool to amplify the capabilities of missionaries. It can help with essential tasks such as effective communication, creating relevant content and training teams, without ever replacing the divine call and mission. As Fei-Fei Li, a Stanford professor and renowned AI researcher, has rightly pointed out, “Artificial intelligence is not a substitute for human intelligence; it is a tool for amplifying human creativity and ingenuity.” (2017).
x. Summary proposal for solving the problem
To overcome the problem presented, we implemented a solution based on Generative Artificial Intelligence (GenAI) that operates locally on personal devices, such as computers. This approach guarantees private and secure interactions, eliminating dependence on unstable internet connections. The solution is designed to process natural language efficiently on the device itself, ensuring that all data and interactions are stored locally, protecting users' privacy. In addition, the ability to interact with local files and documents will allow the AI to provide intelligent and personalized responses, adapted to the context and specific needs of the missions. The balance between efficiency and privacy is crucial, allowing users to take advantage of advances in artificial intelligence while maintaining full control over their information. To achieve this, a careful approach is taken to optimizing the performance of the models, ensuring that they work effectively in environments with limited resources.
x. Clarification note: Theology & Technology
This is a missionary project. However, the explanations follow the project in technical terms, focusing on architecture, functionalities and operational objectives. We will not be using theological or missiological terms. We have chosen this approach to ensure clarity about the infrastructure, security and practical implementation of Generative AI. We understand that questions may arise - we are happy to detail specific aspects or adapt the language to your needs. (RodrigoTinocoBR@gmail.com)
x. The two central technical pillars of the project
x.1. Machine Reasoning Capacity. AI's ability to analyze texts and data, identify patterns and generate logical conclusions, going beyond the mere reproduction of information.
x.2. Natural Language Command. Simplified interface that allows users to interact with AI using everyday language, without requiring technical knowledge.
x. Project objectives in use
- Maintain the local Generative AI solution that works directly on personal devices, without the need for an internet connection.
- Guarantee data privacy and security, ensuring that all interactions and processed information remain on the user's device.
- Expand efficient natural language processing, optimizing the model's performance for environments with limited computing resources.
- Allow interaction with local files and documents, offering intelligent and personalized responses based on the content available on the device.
- Balance efficiency and usability, ensuring that the solution is accessible to users without advanced technical knowledge.
- Optimize the consumption of computing resources, reducing the need for specialized hardware without compromising the quality of the responses generated.
- Enable the use of multiple GenAI models, allowing users to choose between different architectures according to their needs.
- Ensure compatibility with various operating systems, including Windows, macOS and Linux, increasing the accessibility of the solution.
- Eliminate dependence on paid subscriptions, ensuring that users can use the solution without recurring costs.
- Offer template update control, allowing users to choose when and which versions of the templates they wish to use. As well as monitoring by the organization's IT team.
x. Some direct benefits for the missionary field team
Generative artificial intelligence offers several advantages for the missionary field team, especially in the creation and improvement of texts. Simply put, it can contribute in several areas, such as:
✅ Facilitating text production - Writing and improving emails, reports, evangelism materials and other essential documents becomes faster and more efficient, ensuring clarity in communication.
✅ Improving content production - Developing material for blogs, social networks, newsletters and presentations becomes more accessible, saving time and ensuring quality publications.
✅ Structuring training - Creating customized courses, handouts and teaching materials becomes simpler, enabling more effective and well-organized missionary training.
✅ Enriching the missionary's continuing education - Assisting in the search for knowledge through summaries, answers to questions and suggested readings strengthens continuous learning and deepens theological and cultural understanding.
✅ Expanding access to text translation - Translating documents, messages and teaching materials facilitates communication in cross-cultural contexts, allowing for a better connection with different peoples and cultures.
This is an initial list; much more can be done. By optimizing time and resources, generative AI allows missionaries to focus on what really matters: amplifying impact
x. Performance:
The performance of a local AI depends on three main factors: hardware, tool and chosen model. The hardware (CPU, GPU and memory) defines the available processing capacity (personal computer). The tool used to run the AI optimizes execution, influencing speed and efficiency. The model chosen can vary in size and complexity, affecting the quality of responses and resource consumption. For good performance, it is essential to balance these three elements according to the user's needs.
x. Hallucination:
Hallucination in AI, i.e. inaccurate or made-up answers, is a common phenomenon in language models. However, this problem can be reduced by two main factors: empowering the user, allowing them to formulate clearer prompts and critically evaluate the answers, and restricting AI to a local knowledge base, ensuring that the information provided is more accurate and aligned with the organization's thinking.
x. Layers of work required:
x.1 Hardware layer: The project focuses on use on local computers without internet access. We know that the teams have limited equipment. This technology requires good computers. However, we are going to propose minimum equipment for the proposal to run smoothly.
x.2 Software layer: This layer is the software (local backend) responsible for running the Generative AI models directly on the device, for the computing resources (CPU/GPU) and the user interface using natural language (prompt). The software accepts different models.
x.3 Model Layer: GenAI models are architectures trained to process and generate contextualized content, adapting to tasks such as text, data analysis and automation. Both softwares allow you to choose different models in the interface.
x. Learnings and common mistakes in implementation
Generative AI, being an emerging and poorly understood technology, often leads organizations to make strategic mistakes. Generative AI requires an adaptive, non-bureaucratic approach, recognizing its potential as a tool for extending human capabilities, not mere automation. Two critical mistakes stand out in implementations so far:
Underestimating the Collaborative Impact. Treating Generative AI as a centralized tool, similar to traditional systems, ignores its democratic potential. Its power lies in decentralized empowerment: any employee, with the right training, can generate insights, automate tasks or create content. Not adapting processes to encourage this autonomy limits innovation and generates underutilization.
Adopting Single Models. Seeking a universal model ("one-size-fits-all") is a mistake. Each Generative AI model has specific strengths: some specialize in technical text, others in data analysis or creativity. Implementing just one option restricts the solution to a subset of needs and ignores rapid advances in the field.
x. Implementation Plan:
We want to support missionary organizations, aligned with their mission and empowered by Generative AI, capable of expanding their impact without compromising essential resources. For this, a tailor-made plan is crucial, taking into account the operational reality, local challenges and specific objectives of each small missionary organization. The central reflection should be: how to structure a suggestive, non-prescriptive journey that guides these organizations to adopt AI gradually, ethically and adapted to their unique context - whether in the translation of materials, analysis of community data or engagement of volunteers.
x. Validations already carried out
Pilot Project 1
Hardware: MacBook Pro M3 Pro
Software: Gpt4all: https://www.nomic.ai/gpt4all
Model: Reasoner v1 (based on "Qwen2.5-Coder 7B")
Perfomace: Excellent
Result: very good
Pilot Project 2
Hardware: MacBook Pro M3 Pro
Software: LM Studio: https://lmstudio.ai
Model: DeepSeek R1 distilled into Llama 8B
Perfomace: low
Result: very good
Note: Software and model are free products. You can do your own tests. Tell us about your experience.
x. Complementary projects
In this project, we are focusing on the use of Generative AI on local machines, without the need for an internet connection. However, for effective implementation, two other projects are essential: training in the solution and building a knowledge base. Both will be dealt with separately. Below is a brief description of each:
x.1 Private Knowledge Base. A collection of documents, data and information unique to the organization, such as religious texts, records of communities served and internal reports. This base will allow interaction with local files, providing intelligent and personalized responses based on the content stored on the device.
x. 2. GenAI Enablement. To maximize the use of Generative AI, it is essential to train different audiences: Technical Teams - They must understand optimization models, tools and practices to ensure efficiency and security. End Users - They need to learn how to create effective prompts, since the quality of responses depends on well-structured commands. Leadership - They need to understand the strategic impact of AI in order to guide its proper use. Creating learning paths for each audience will ensure greater productivity, minimize errors and allow everyone to take advantage of AI's potential consciously and effectively.
x. Future projects
Generative AI is still in its infancy. The more organizations and their members understand its potential, the greater its application will be. This project focuses on text production, but its possibilities go further. Expansion will occur as the technology is consciously and effectively adopted by the organization.
x. Implementation schedule
The estimated time for implementation is 4 weeks.
x. Cost - how to invest
We are excited to share that the entire project will be made available free of charge to missionary organizations, allowing more people to benefit without financial barriers. However, it is important to note that the necessary equipment, such as personal computers, will be the responsibility of each missionary.
The project is already underway. We can expand. We are currently seeking an initial investment of $2.000 (two thousand dollars) to purchase laptops that will be used for compatibility tests with different hardware. These tests are essential to ensure that the technological solution meets the specific needs of the missions in various conditions.
We invite you to be part of this mission in several ways. Firstly, through your prayers, which are fundamental to our work. Secondly, through financial support, which will enable us to acquire the resources needed to develop and implement the project. We work as dedicated volunteers, and the livelihood of our families and the continuity of our work depend on the generosity of organizations, churches and individuals who share this vision.
Your contribution is vital so that we can continue to serve and expand our missionary impact through innovative initiatives like this one. We ask that you pray about how you can support us and consider making a donation in whatever way and for however long you feel like it. Together, we can transform lives and strengthen mission work around the world.
To contribute:
📌 In Brazil - Via CEL-PIX: 61-9-9966-4046
🌍 Abroad - Via PayPal: rodrigotinocobr@icloud.com
USA – USA - We have a tax deduction option (ask for more information)
If you prefer other ways of contributing, we're happy to help. Together, we can unite faith, technology and missionary impact where it's needed most.
x. Final Words:
This project is innovative and necessary, especially for its focus on democratizing Generative AI for remote missionary contexts, where the technology is underused but the potential for impact is immense.
If you'd like to explore technical details, scalability strategies or adjustments for specific cases, I'm happy to contribute: RodrigoTinocoBR@gmail.com.
x. Author's presentation
I'm Rodrigo Tinoco, a full-time cross-cultural missionary and digital technology specialist, with certifications from Microsoft, IBM, Cisco and others since 2004. I work as a solutions integrator, identifying the needs of missionary organizations and implementing high-performance technologies to boost their impact. In the projects I collaborate on, my focus is on uniting technology and mission, providing tools that optimize communication, management and access to information. All my work is voluntary, serving to strengthen missionary initiatives and broaden their reach.
x. Glossary:
This glossary brings together essential terms to facilitate the understanding and practical application of Generative AI in the context of small organizations. It is not complete; we will expand it with each version. It is part of a proposal for literacy in this technology.
Generative AI (GenAI):
Technology capable of creating texts, images or analyses from patterns learned from data. Example: Generating automatic reports based on field data.
LLM (Large Language Model):
Large-scale language models, with billions of parameters, used for complex tasks. Example: Translating religious texts into minority languages.
SLM (Small Language Model):
Compact and efficient models, ideal for local devices. Example: Automating responses to donors on a basic laptop.
RAG (Retrieval-Augmented Generation):
Technique that combines text generation with local document search for precise answers. Example: Answering questions using the organization's internal manuals.
Prompt:
Command in natural language that directs the AI to perform a task. Example: "Summarize the March activity report in 3 paragraphs."
Hallucination:
Phenomenon in which the AI generates incorrect or invented information. Mitigation: Use RAG and human review.
Model Repository:
Local library of AI models, managed by the IT team. Objective: Version control and updates.
Missionary organizations
majority world
Comments