The Problem with Synthetic Data Isn’t Bias. It’s Average.

In The AI Journal: the real problem with synthetic data isn't bias, it's that it drifts to the average and misses real people.