Datagen is a company that generates synthetic data for AI training. While it offers unlimited amounts of data, the quality of the data can be questionable. The data-generation process can lose nuances and intricacies in real data, and the photorealism capabilities of the generated faces are unlikely to fool a person into thinking they're real. However, the realism of the data is not crucial for every application. Synthetic data is not a silver bullet for AI bias, as any biases present in the data-generation process will be picked up by models trained on it. A study by Arizona State University showed that when trained on predominantly white, male images of engineering professors, its generative model amplified the biases in the dataset, meaning that it produced images of minority modes less frequently.
