Forum Discussion
Iceberg Ahead: The Unseen Data Divide in AI
85% of AI's training data comes from the global north, highlighting the digitalization gap of the global south. This imbalance is more than just a discrepancy; it represents a profound division that ...
The imbalance in training data is a real concern when it comes to global equity. What I think will be best is supporting local data collection and developing smaller models that can work offline.
Federated learning can also help keep sensitive information local. Having diverse development teams and taking a 360-degree view of the problem can help identify bias much earlier.
The outcome usually is AI that reflects a wider range of human experiences. Local partners should also have a meaningful say in how these systems are developed and used.