The generation of 3D conformations for organic molecules is of critical importance in ligand docking, 3D pharmocophore, and other applications in molecular modeling and computer-aided drug design. Traditionally, two types of methods have been applied to sample the conformational space, involving either the systematic enumeration of torsional angles or the stochastic perturbation of existing conformations. In recent years, new techniques based on self-organization emerged as an alternative solution to the conformational sampling problem, in which the atomic coordinates of the molecule are assigned random initial values and then gradually refined toward a feasible geometry. Stochastic proximity embedding (SPE) is the first algorithm using such an approach. SPE refines the geometry by repeatedly adjusting the pair-wise distance between two randomly selected atoms. Recently, we proposed a self-organizing superimposition (SOS) algorithm, where precomputed conformations of rigid fragments are used as templates to enforce locally optimal geometry. Compared to SPE, the SOS algorithm was shown to generate conformers of better quality at a faster speed. The SPE and SOS algorithms can be applied to any molecule, including those containing many rotatable bonds or macro-cycles, which are more challenging to traditional methods. They also promise potential applications on more complicated problems such as the NMR structure determination of proteins.