I’ve worked on and adjacent to systems using synthetic data and can give you an example from the computer vision world. We had a model that was trained on video of human faces that we collected and paid participants for. For numerous non-tech reasons, this meant we couldn’t collect as much data on

People of color to accurately understand their faces, leading to a bad bias we wanted to correct. One of the techniques we used was GANs for generating faces that we could train our models on. Our models suddenly were much more accurate on PoC. This is because the data is good enough for the model