ThisMultiparty computation chapter introduces the concept of multiparty computationMultiparty computation and discusses some simple examples that serve as inspirations for the development of federated learningFederated learning. The idea of multiparty computationMultiparty computation is to have many parties, each of whom has only a small data set, agree to pool together their data resources to conduct analysis that would otherwise be impossible or lead to results having no practical value. In the most basic multiparty computationMultiparty computation framework, each participating party sends encrypted, raw data to an external trusted authority for data aggregation, management, and analysis. In some circumstances, all parties may be unable to agree on a suitable external authority. In view of this difficulty, some multiparty computationMultiparty computation frameworks without a trusted authority have been developed, but they involve generation and processing of more data and more computational work. These multiparty computationMultiparty computation frameworks are technically viable solutions for small data sets. However, for applications involving training of AIArtificial intelligence (AI) models, they would involve encryption, transmission, processing, and decryption of such large volumes of data and would be riddled with so many legal business and technical challenges that real-world deployment would be impossible.