Nine years ago, Google CEO, Sundar Pichai, talk say artificial intelligence (AI) go make information dey “universally accessible” to everybody, no matter di language wey dem dey speak.

Since dat time, e don dey repeat dis promise, make people for di world dey hope say technology go help bridge di gap wey dey between different languages and give everybody equal access to knowledge.

But for Africa, wey get more than 2,000 languages, dis promise still dey far. Plenty people for di continent still dey find am hard to use di advanced AI tools wey dey change agriculture, education, and everyday life, because di tools no sabi or fit communicate for dia language.

Research show say ChatGPT, wey get 800 million active users every week for di world, fit only understand 10 to 20 percent of sentences wey dem write for Hausa, one language wey over 94 million Nigerians dey speak.

Di same thing dey happen for other African languages like Yoruba, Igbo, Swahili, and Somali, even though millions of people dey speak dem. But di question be say, why AI tools no dey focus on African languages and wetin dis one mean for who dey shape di digital future?

One big reason na di “low-resource” problem. Dis one mean say, di materials wey dey online for African languages, like websites, books, and transcripts, no plenty. AI tools like ChatGPT dey rely on plenty digital data to learn, and most of di data dey for English or other big languages for di West.

Hellina Hailu Nigatu, wey dey do research for low-resource languages for University of California, Berkeley, talk say di way dem dey measure progress for AI dey based on wetin dey work for Western languages. Dis one dey make AI tools struggle to understand or generate African languages, no matter how many people dey speak dem.

Another reason na di priorities of di global AI companies. Most of di standards wey dem set dey favour Western languages wey get plenty online materials and funding. African languages no dey get di same attention because di companies no see immediate money wey dem fit make from dem.

Even when African languages dey inside di AI system, di tools dey sometimes misrepresent di local culture or dey carry Western stereotypes. Dis one dey show di bias wey dey di system and di inequality for who dey shape di AI tools.

Another wahala na di technical challenge. Research show say di cost to process text for languages wey no use Latin alphabet dey higher. Dis one mean say people wey no too get money go dey pay more to use di tools and di results no go even dey accurate sometimes.

But hope dey. One big project wey dem call African Next Voices don start to change di story. Di project, wey Gates Foundation dey fund with $2.2 million, dey create AI-ready language data for African languages. Researchers for di continent don record 9,000 hours of speech for 18 languages across Nigeria, Kenya, and South Africa.

Ife Adebara, wey be chief technology officer for Data Science Nigeria, talk say di project go help improve di way AI dey handle languages like Hausa, Yoruba, Igbo, and Naija. Di researchers dey work directly with communities to collect authentic language data, instead of just scraping di internet like Western companies dey do.

For Kenya, Lilian Wanzare, wey dey Maseno University, dey lead di effort to collect data for languages like Dholuo, Kikuyu, Kalenjins, Maasai, and Somali. Her team dey ask people to describe wetin dem see for pictures with dia native language, so di data go dey real and local.

For South Africa, Vukosi Marivate from University of Pretoria dey work on seven languages like Setswana, isiZulu, isiXhosa, and others. Di goal na to create AI language models wey businesses fit use to improve dia tools.

Dis project dey show say African languages no be afterthought. Di researchers dey prove say di future of AI fit dey shaped by Africans, for Africans. Organisations like Masakhane don already build strong networks for natural language processing, and dis one dey show wetin fit happen when Africans take charge of dia own digital destiny.