Black founders have observed that mainstream large language models like ChatGPT exhibit significant shortcomings in handling cultural differences. These models often provide overly generic responses when addressing questions unrelated to Western cultural contexts, lacking an understanding of African Americans and their unique cultural heritage.
To address this issue, over the past year, many black founders have launched alternative ChatGPT models tailored for communities of color. John Pasmore's Latimer.AI is specifically designed for African Americans; Erin Reddick's ChatBlackGPT will introduce a similar product in June; and Spark Plug's AI in Canada focuses on providing native English support to black students.
These founders have expressed that mainstream AI companies overlook minority cultures when building models, resulting in algorithms with a pronounced Eurocentric bias. To rectify this, they have collected and trained on extensive cultural materials from their own communities, ensuring that the product outputs are more aligned with the actual experiences of black communities.
In addition to language models, renowned material companies like pocstock are also striving to provide more diverse training data, avoiding the over-monolithic representation of races in AI-generated images. Some African entrepreneurs are also focusing on incorporating over 2,000 languages and dialects from the region into large AI models to ensure a comprehensive understanding of African culture.
These innovations are helping to alleviate the long-standing cultural deficiencies in mainstream AI models. Some analyses suggest that inclusive AI could become a trillion-dollar market in the future, and there is anticipation for more entrepreneurs of color to contribute in this field.