Frequently Asked Questions
Everything You Need to Know About Contributing
Click any question to see the answer. Can't find what you're looking for?
Contact UsAbout Contributing
A: Absolutely YES!
Here's the math: 10 people with 5 documents each = 50 documents. 50 documents × 10 pages each = 500 pages. That's significant!
Plus, you might know where to find MORE materials later, or connect us with others who have materials. Every single contribution matters.
A: Of course! If you can do these three things, you can contribute:
- Use Google Drive (like Gmail or WhatsApp)
- Take a photo with your phone
- Upload files (click and drag)
That's it! All the complex technical work (AI training, data processing, quality checking) is handled by the technical team. We've designed the system specifically for non-technical users.
A: YES! We have technology specifically designed to handle:
- ✅ Faded text (we enhance contrast)
- ✅ Torn or damaged pages (we can fill gaps)
- ✅ Water damage or stains
- ✅ Old yellowed paper
- ✅ Handwritten text
- ✅ Old typewriter text
The rule: If a human can read it, we can probably digitize it. Even if you're not sure, share it and we'll assess.
A: No problem at all!
We have an "Unsorted" folder specifically for materials you're uncertain about. The technical team reviews everything regularly and moves files to the correct tier.
You can't really make a mistake that causes problems. Worst case: A document sits in "Unsorted" for a few days until we review and categorize it.
A: Very flexible! Here's typical time investment:
- Initial Setup: 5-10 minutes (get access, understand system)
- Finding Materials: Varies (30 min to ongoing)
- Uploading: 2-5 minutes per item
Total: Minimum 1-2 hours (one-time), typical 5-10 hours spread over weeks/months. You work at your own pace - no deadlines, no pressure!
Privacy & Ownership
A: You have complete control over privacy. Options include:
- Fully public: Can be shared openly (most historical/cultural materials)
- Project-only: Used for training, not publicly shared
- Anonymous: Your name not associated with it
- Restricted: Specific limitations on use
You decide for each item. Before any public release, we get explicit permission and respect any restrictions you set.
A: Legal ownership depends on the source:
You keep ownership if:
- You created the materials
- They're your personal writings
- You hold the copyright
What the project gets:
- Right to USE for AI training
- Right to DIGITIZE and preserve
- Right to SHARE (with your permission)
- Attribution to you as contributor
Analogy: Like donating books to a library - you're sharing access while respecting original ownership.
A: You can withdraw materials anytime:
Process: Contact us → Specify what to remove → We remove it from future use
Important notes:
- We CAN remove from future training
- We CANNOT unlearn from past training (technically impossible)
- Already trained AI versions keep what they learned
- No one will have access to your actual files
Think of it like publishing: You can stop distributing a book, but people who already read it remember the content.
About the Project
A: Google Translate treats Gĩkũyũ as "English with different words." We don't.
Google Translate Problems:
- ❌ Learns from mostly English data
- ❌ Misses cultural context in proverbs
- ❌ Confuses ũ/u and ĩ/i (changes meaning!)
- ❌ Robotic, unnatural phrasing
Digital Elder Advantages:
- ✅ Learns pure Gĩkũyũ first (thinks like a native speaker)
- ✅ Understands cultural context
- ✅ Preserves critical language features
- ✅ Natural, authentic phrasing
A: 4-5 months from now to working prototype:
- Now - Jan 2025: Data collection (gathering materials)
- Feb-Mar 2025: Phase 1 training (pure Gĩkũyũ)
- April 2025: Phase 2 training (translation capability)
- May 2025: Beta testing with contributors
This is realistic for 10-15 hours/week of work plus community collaboration.
A: Under $400 total for the entire project!
Breakdown:
- Data Collection: $5-10/month (Google Drive storage)
- Phase 1 Training: $100-200 one-time (GPU rental)
- Phase 2 Training: $50-100 one-time (GPU rental)
This proves language preservation AI is achievable without massive funding. Contributors don't pay anything - it's all volunteer.
A: YES for community and educational use!
Free Access:
- Personal translation needs
- Learning Gĩkũyũ
- Educational institutions
- Community organizations
- Contributors get lifetime access
Potential Paid Use:
- Commercial translation services (businesses)
- Large-scale API access
- Revenue supports ongoing development
A: Even if AI training fails, valuable work is preserved:
Guaranteed Outcomes:
- ✅ Digital archive of Gĩkũyũ materials created
- ✅ Your materials preserved and backed up
- ✅ Foundation for future attempts established
- ✅ Community connections strengthened
Worst case: Materials are preserved digitally, and we've learned valuable lessons for next attempt.
Best case: We create the first African language AI of its kind!
Technical Questions (Simplified)
A: Think of it like teaching a very smart but initially ignorant student:
Traditional Programming: Programmer writes exact rules ("If X, do Y")
AI/Machine Learning: Show the AI thousands of examples, it finds patterns on its own
Our Approach:
- Feed AI lots of pure Gĩkũyũ text
- It learns grammar, patterns, cultural logic
- Then learns to translate
- Finally can answer questions and translate
Analogy: Like a child learning language - first learns mother tongue naturally, then learns second language, eventually can translate between them.
A: Because these marks completely change meaning in Gĩkũyũ!
Examples:
- mũrĩ (year/age) vs muri (you are)
- gũkũ (here) vs guku (chicken)
- mũndũ (person) vs mundu (nothing)
If the AI learns wrong patterns, it mixes up completely different words, creates nonsense sentences, and loses all meaning.
Our solution: Quality checking by native speakers, verification of critical texts, prioritizing sources with diacritics, manual correction when needed.
Bottom line: Better to have less text that's correct than lots of text with errors.
A: Teaching order profoundly affects what AI learns.
Wrong Approach (Most Current Systems):
- Mix everything together - Gĩkũyũ + English all at once
- AI learns to rely on English patterns
- Result: Translation, not understanding
Our Approach (Curriculum Learning):
- First: Pure Gĩkũyũ only → AI learns to "think" in Gĩkũyũ
- Second: Add English connections → AI learns to explain in English
Analogy: Teaching a child English and Gĩkũyũ simultaneously vs child learns Gĩkũyũ first (mother tongue), becomes fluent, THEN learns English.
Research shows this approach creates more authentic language understanding.
What You Get
A: Direct benefits for future generations:
Language Learning:
- Tool to help children learn Gĩkũyũ
- Interactive practice (not just books)
- Cultural context built in
- Makes learning engaging
Cultural Connection:
- Access to proverbs and their meanings
- Traditional stories preserved
- Elder wisdom available
- Heritage maintained
Practical Use:
- Homework help in Gĩkũyũ
- Translation assistance
- Communication across generations
A: Absolutely! That's part of the vision.
Gĩkũyũ is the PILOT PROJECT:
- Proves the approach works
- Creates replicable process
- Documents challenges and solutions
- Builds templates and tools
Can be adapted for:
- Kiswahili (obvious next step)
- Luo, Kamba, Kalenjin (Kenya)
- Yoruba, Igbo, Hausa (Nigeria)
- Any language with dedicated community
Your contribution helps preserve ALL African languages, not just Gĩkũyũ!
Still Have Questions?
We're here to help! Contact us directly and we'll answer any questions you have.