Then they are simply making it worse by their utter incompetence.
It is a nice idea but too expensive/impractical to give useful results - especially given our limited test capacity.
Also, though most models treat the epidemic as a stochastic process, in reality infections and contacts don't pop up randomly but exist within networks. So the sampling would need a quite sophisticated design, relying on information we probably don't have to hand.
ETA: There is an app available in which anyone/everyone can record their symptoms or non-symptoms daily. If everyone used that then we could start to move towards a good idea of how many people were infected with symptoms. Without symptoms is still a problem.