If OpenAI can accidentally train its flagship model to obsess over goblins, what other more subtle and potentially harmful biases are being reinforced through the same feedback loops?
The keynote sound bite that everybody in the world could be a programmer is now a reality that people are living.