Constructionism, long before it had a name, was intimately tied to the field of Artificial Intelligence. Soon after the birth of Logo at BBN, Seymour Papert set up the Logo Group as part of the MIT AI Lab. Logo was based upon Lisp, the first prominent AI programming language. Many early Logo activities involved natural language processing, robotics, artificial game players, and generating poetry, art, and music. In the 1970s researchers explored enhancements to Logo to support AI programming by children. In the 1980s the Prolog community, inspired by Logo's successes, began exploring how to adapt logic programming for use by school children. While there have been over 40 years of active AI research in creating intelligent tutoring systems, there was little AI-flavoured constructionism after the 1980s until about 2017 when suddenly a great deal of activity started. Amongst those activities were attempts to enhance Scratch, Snap!, and MIT App Inventor with new blocks for speech synthesis, speech recognition, image recognition, and the use of pre-trained deep learning models. The Snap! enhancements also include support for word embeddings, as well as blocks to enable learners to create, train, and use deep neural networks. Student and teacher project-oriented resources highlighting these new AI programming components appeared at the same time. In this paper, we review this history, providing a unique perspective on AI developments--both social and technical--from a constructionist perspective. Reflecting on these, we close with speculations about possible futures for AI and constructionism.