You have to start with a heterogeneous seed of a different material. If possible as close to the silicon structure as possible (not sure what is used in Si growth). You then hope for a monocrystalline part and prepare that as a seed for the next growth experiment. And after that (if successful) you start searching for the suitable crystallographic directions :-)
From my limited experience in growing complex ceramic oxides using the floating zone furnace we start with 2 pressed powdered rods. This means they are made up of randomly orientated crystals to start with. In the FZ furnace the lower powdered rod acts as the seed and whilst it is not a single crystal the tip of the top rod can be melted in a controlled manner. If the melted tip of the top rod is placed on the bottom seed rod in a controlled manner, i.e. it isn't dripped resulting in rapid cooling, then a single crystal can be grown from the newly melted - solidified crystal upwards. Getting good crystals is all about the rate of solidifying the melted tip.