The equilibrium population of different conformational states of a polypeptide can in principle be obtained by a very long molecular dynamics simulation. The method of internal coordinate molecular dynamics earlier developed in this laboratory (A.K. Mazur and R.A. Abagyan J. Biomol. Struct. Dyn. 6,833 (1989)) allows one to use time steps much larger than usual for computing molecular trajectories. It is shown here that the sampling of the conformational space can be additionally enhanced by adding a random component to the set of forces applied to atoms. We describe the algorithms by which the random force is introduced and also a special method which excludes the fast rotation of polar hydrogens from equations of motion but keeps them movable. As a result the task stated in the title becomes realistic. Internal coordinate stochastic dynamics is applied for scanning the conformational space of the pentapeptide Met5-enkephalin which is a common test example widely used in theoretical studies. A large number of conformational transitions is observed during the 20 ns simulation starting from the global energy minimum thus allowing us to arrive at a nearly Boltzmann distribution of populations of conformational states. A few states are found which are distinguished by high apparent configurational entropy which turn out to correspond well to experimentally observed conformations of enkephalins.