In this expository lecture, we explain how problems of classical polynomial algebra may be understood and numerically solved when they appear in the context of Scientific Computing, i.e. with data of limited accuracy. A suitable conceptual framework is introduced and its use explained in a number of situations.