3 Challenges Threaten Africa Rise of AI in 1000 Languages! Can They Solve It?

Nigeria is developing a multilingual AI model to improve digital inclusion, focusing on five local languages and accented English. Many African languages, like Hausa and Swahili, lack sufficient datasets for AI training, limiting their presence in global AI systems. Startups like Jacaranda Health and Lelapa AI are addressing this gap with tools like UlizaLlama and VulaVula for healthcare and translation. However, challenges such as data scarcity, ethical concerns, and lack of regulations hinder progress. Experts emphasize the need for fair compensation and proper frameworks to protect language communities. Despite these hurdles, Africa’s AI ecosystem is growing, aiming for a more inclusive digital future.

3 Challenges Threaten Africa Rise of AI in 1000 Languages! Can They Solve It?

When Nigeria announced plans to develop a multilingual artificial intelligence (AI) tool in April, it was exciting news for many in the tech community. Among them was 28-year-old Lwasinam Lenham Dilli, a computer science student who had struggled to find the right datasets for his final-year university project. His goal was to create an AI-powered chatbot that could communicate in Hausa, one of Nigeria’s most widely spoken languages.

“I needed English texts alongside their Hausa translations, but finding clean and usable data online was almost impossible,” Dilli told the Thomson Reuters Foundation.

Why African Languages Matter in AI

Dilli’s challenge reflects a broader issue facing AI development in Africa. While tools like OpenAI’s ChatGPT, Meta’s Llama 2, and Mistral AI have revolutionized the tech landscape, they often fail to handle African languages properly. Many AI models struggle to understand or generate meaningful content in languages like Hausa, Amharic, or Kinyarwanda, making them less useful to millions of people across the continent.

Technology experts warn that without AI models tailored for African languages, digital access and economic opportunities could become even more unequal, leaving many people behind in the AI-driven future.

Nigeria’s Plan for an Inclusive AI Tool

To address this gap, Nigeria’s Digital Economy Minister Bosun Tijani announced an initiative to develop a multilingual large language model (LLM) that would better represent the country’s linguistic diversity. This AI model will be trained on five Nigerian languages—Yoruba, Hausa, Igbo, Ibibio, and Pidgin—as well as accented English to ensure it captures the way people naturally speak.

The project will rely on contributions from local AI startups and volunteers fluent in these languages. Additionally, it will tap into the expertise of over 7,000 participants from Nigeria’s tech talent program, which aims to train three million individuals in programming and coding.

The Complexities of Building African AI

Silas Adekunle, co-founder of the AI startup Awarri, emphasized that building an AI model that understands Nigeria’s diverse linguistic and cultural landscape is no easy feat.

“With Nigeria’s many accents and languages, developing a large language model (LLM) comes with serious challenges,” Adekunle explained. “But if we get it right, this project could empower developers and businesses to create AI-driven solutions tailored for Nigeria.”

Given the limited resources available, the team has had to be creative in collecting data, training the model, and ensuring accurate language representation.

Africa’s AI Boom in Local Languages

Africa is home to over 2,000 languages across 54 countries, according to UNESCO. Yet, these languages remain underrepresented online, with English dominating around half of all websites, followed by Spanish, German, Japanese, and French.

While Nigeria’s government-led AI initiative is a step forward, several African startups are also working to fill the gap by developing AI tools for languages like Swahili, Amharic, Zulu, and Sesotho.

For example, in Kenya, health-tech company Jacaranda Health recently introduced UlizaLlama (which means “Ask Llama”), an AI model built on Meta’s Llama 3 system. Designed to improve maternal healthcare in East Africa, UlizaLlama provides Swahili-speaking expectant mothers with instant answers to questions about pregnancy, such as dietary guidelines and fetal movement. By the end of June, the platform aims to personalize responses based on individual needs, offering more tailored support.

African AI for Health and Translation

Jay Patel, director of technology at Jacaranda Health, explained that UlizaLlama aims to provide accurate and fast responses to expectant mothers who may lack easy access to information. Initially targeting an 85% accuracy rate, the AI hopes to cut response times from a few minutes to under a minute in the future.

In South Africa, the Masakhane initiative is using open-source machine learning to improve African language translation. Meanwhile, Lelapa AI, a South African research lab, has developed VulaVula, a commercial tool capable of translating, transcribing, and analyzing languages such as English, Afrikaans, Zulu, and Sesotho.

The Data and Ethics Dilemma in African AI

Despite these advancements, African AI development faces significant roadblocks. One major challenge is data scarcity. Many African languages are classified as “low-resource” because there isn’t enough high-quality data available to train AI models effectively.

Another concern is ethics. Collecting language data from African communities raises issues of consent, compensation, and copyright. Many African societies rely on oral traditions, and some communities may be hesitant to share language data without clear protections in place.

Michael Michie, co-founder of Everse Technology Africa, pointed out that most African countries lack regulations on AI data collection, meaning communities risk being exploited if their language contributions aren’t fairly recognized or compensated.

While open-source initiatives like Creative Commons can help with data sharing, they don’t fully address concerns about fair use and compensation. Vukosi Marivate, an AI researcher at the University of Pretoria and co-founder of Lelapa AI, warned against blindly applying open-source frameworks without proper safeguards.

“We need to make sure African languages and the people who speak them aren’t left out of AI development. The goal should be to uplift communities, not exploit them,” Marivate emphasized.

The Future of AI in Africa

As investment in AI increases across Africa, there is growing interest in making AI tools more inclusive. However, without better data resources and ethical protections, the risk remains that Africa’s rich linguistic diversity could be overlooked in the global AI race.

The success of initiatives like Nigeria’s multilingual AI model and UlizaLlama will depend on how well they navigate these challenges. If done right, they could mark the beginning of a more inclusive AI future—one where African languages and cultures are fully represented in the digital world.