Multi-component aluminum-based alloys are important in various industries. The percentage of the base component with various impurities and alloying additions in such alloys plays a fundamental role in achieving the required properties of the final product. For our research, information on the diffusion properties of alloying components, such as Si, Cu, Mg, Sc, Ca, Ti, Zr, Cr, in the liquid state is useful. Unfortunately, for these elements (with the exception of Cu), even in the pure form, the required information in the literature is insufficient. In this work, the self-diffusion coefficient of liquid Mg is chosen as the object of study. The calculations are carried out, starting from the melting temperature with a step of 30 K, in the range of 923–1043 K within the framework of an approach based on three theoretical components—the linear trajectory method, the model square-well potential, and the random phase approximation. The results obtained are comparable with all the data of a computer experiment available in the literature, although to a better extent with the result of classical MD. At the same time, a slight increase in our calculated values of the self-diffusion coefficient with temperature may promise their better agreement with the ab initio MD results with a significant increase in temperature, which is due to the increase in RPA accuracy with increasing temperature. The study shows that in the absence of experimental information on the self-diffusion coefficient of a liquid metal, the approach of joint use of the linear trajectory method, the square-well potential, and the random phase approximation can be a workable tool for obtaining estimated values of this property.