The quantized representation of signals is a general task of data processing. For lossy data compression the celebrated Rate-Distortion theory provides the compression rate in order to quantize a signal without exceeding a given distortion measure. Recently, with the Information Bottleneck method an alternative approach has been emerged in the field of machine learning and has been successfully applied for data processing. The fundamental idea is to include the original source into the problem setup when quantizing an observation variable and to use strictly information theoretic measures to design the quantizer. This paper introduces this framework and discusses algorithmic implementations for the quantizer design.