When querying AI models like ChatGPT for African-origin content, the responses are often vague and lack depth.

This is a result of the models’ limited training on African datasets, which leaves them without a sophisticated grasp of the linguistic and cultural backgrounds of the continent. 

Read also: Plaud AI Voice Recorder: AI tool for superfast audio to text transcription

The need for African Data curation

To address this, panellists at the Moonshot event organised by TechCabal stressed that African countries should gather and share their datasets online. 

Data curation, however, has challenges, such as high costs of Internet services when uploading data and low digital literacy levels in the continent. It has been noted that certain African languages and cultures require more textual resources, which makes it challenging to develop data sets.

Bayo Adekanmbi, Founder of Data Science Nigeria, proposed using voice-to-text technology to document local languages and cultural nuances, while Nigerian startup Intron Health is leveraging voice-to-text for medical records by allowing healthcare professionals to input data through speech.

Incorporating code-switching in AI models

In many African communities, languages like Pidgin and Yoruba are commonly used in speech patterns. 

AI models can more accurately simulate real-world conversations by incorporating code-switching. 

Mr Adekanmbi recommended that startups should develop AI models that understand and process these unique linguistic contexts.

Read also: OpenAI streamlines AI voice assistant development

Public-private partnerships for improved AI adoption

Lavina Ramkisson, an AI Board member at GSMA, and Olumide Okubadejo, the Head of Product at Sabi, said that global partnerships are not only beneficial but crucial for enhancing data documentation and AI development in Africa. 

It is imperative that the public and private sectors collaborate to improve infrastructure, skills, and data accessibility in order to fast-track the adoption of AI.

Addressing Africa’s data scarcity issue is not a simple task. It requires a multi-faceted approach involving strategic partnerships, contextual AI model building, and cutting-edge  technology.

African nations may increase the impact of AI and develop inclusive technologies that take into account the continent’s varied linguistic and cultural terrain by working together and implementing a comprehensive plan.