De-colonial AI a cll for help and guidance and even assistance

Community
1

LinkedIn Post: Call for Collaboration on Decolonizing AI Models

:earth_africa: Seeking Collaborators: Help Develop Afrocentric and Anti-Colonial AI Models

As artificial intelligence continues to shape the future, the majority of AI systems and models still reflect Eurocentric biases, often sidelining African knowledge systems, histories, and philosophies. While there are excellent models focused on African languages, like AfroXLMR and SERENGETI, there is a critical need for AI that truly embodies Afrocentric perspectives and anti-colonial ideologies.

I am working on an initiative to develop an Afrocentric, anti-colonial AI model, and I’m seeking assistance in any of the following areas:

:books: Data Curation: Help curate datasets that reflect African post-colonial thought, indigenous knowledge systems, and anti-colonial literature.

:handshake: Collaborations: I’m looking to partner with African scholars, AI experts, and institutions interested in developing AI models that challenge colonial narratives and promote African worldviews.

:hammer_and_wrench: Technical Expertise: If you have experience in AI development, NLP, or machine learning, your expertise could help fine-tune models like GPT-3 or AfroXLMR to reflect Afrocentric philosophies and perspectives.

This is more than a technical project—it’s about reshaping AI to represent diverse global perspectives and move away from the historical dominance of Western narratives in technology.

:mailbox_with_mail: If you’re interested in collaborating or can offer guidance, resources, or connections, I would love to hear from you. Let’s work together to create AI that reflects the richness and complexity of African thought and decolonized knowledge.

Feel free to reach out or comment below if you’d like to join this important journey.

ai #MachineLearning nlp #AfrocentricAI #DecolonizingAI #TechForGood #Collaboration #AfricanPhilosophy #IndigenousKnowledge #AIforAfrica

2

We often talk about trying to make models that are unbiased but this makes me wonder if there isn’t value to having models that are heavily biased towards a particular world view or ideology.

One issue with current models is that they all seem to be converging on roughly the same neutral and often boring view of the world. One of the things that makes the world interesting is its diversity of view points. It seems like we could stand for a little more diversity with models.

One approach would be a spin on RLHF which leverages groups of like minded individuals to tune a given model to better align with their shared view. It seems like you could just work as a group to build a rich fine tuning dataset that could then be applied to any model.

3

I think “decolonizing” British English of the legacy of Roman, Viking, Anglo-Saxon, Danish or French colonization would also be an interesting challenge. Removing words and names derived from Latin, Norse, Danish and French might be a challenge let alone any foundational legacies like civic centres, sewers, farming techniques, highways, aqueducts or even philosophies … the list goes on …

Tough to disentangle I think.

4

Maybe calling it African Centric could be a little off putting to some thinkers, perhaps New World - Old World centred, AI is a powerful tool so im sure something could made to happen…