The dataset will be used in a competition to detect genuine vs computer generatd voice. So presumably in the future they can draw inspiration from the winning a

The dataset will be used in a competition to detect genuine vs computer generatd voice. So presumably in the future they can draw inspiration from the winning approach and integrate it as the adversarial side of the GAN used to synthesize voices. Rinse repeat, each generation of state of the art fake detection feeding the next generation of creation. I can see this escalating past human levels of detection.