How relevant are Marr's three levels in modern vision science?