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?

Pretty much all the movies you can catch in the Twin Cities this week.